The Napkin Token

The Problem: The IPO Was Never Built for You

Start with the census. Roughly 377 million businesses exist worldwide (Statista, 2023) [1], and that count is conservative. Fewer than 50,000 are publicly listed (World Bank/WFE, 2024) [2]. Run the division: about 0.013%, or 1 in 7,700. Public markets (the greatest wealth-creation technology ever built) serve a rounding error of the world's companies.

The companies locked out are not marginal. Small and medium enterprises are roughly 90% of all businesses and 60-70% of employment globally (UN/World Bank/ILO) [4]: well over 2 billion workers, roughly two of every three jobs on Earth [4]. They produce about half of global GDP (UN) [4], and in emerging markets they create seven of every ten formal jobs (IFC/World Bank) [5]. And the capital system fails them at measured, trillion-dollar scale: the IFC now puts the formal SME financing gap at $5.7 trillion, up 27% since 2015, and roughly $8 trillion counting informal firms (IFC/SME Finance Forum, March 2025) [5]. The gap is compounding at roughly 6% a year [5]. The system fails precisely where the jobs are created. This is not founder rhetoric. It is the World Bank Group's own arithmetic.

The public market is shrinking, not growing

Here is the part the exchanges would rather not discuss. The United States had about 8,000 listed companies at the 1996 peak. By end-2024 it had roughly 3,950 (CRSP data via Apollo/Meketa) [6]. A halving, while the economy tripled. The Wilshire 5000, an index built to hold 5,000 stocks, holds about 3,600 (Wilshire, 2025) [6]. The American market literally cannot fill its own index. In 1996 the US produced 677 real operating-company IPOs; from 2022 to 2025 it averaged about 64 a year (Jay Ritter, University of Florida, Dec 2025) [7]. Globally, H1 2025 saw just 570 IPOs, down 20.9% year over year (WFE) [2]. The front door is narrowing while the queue outside grows by millions of companies a year.

The toll booth

Why? Because the IPO is priced like a luxury good. Underwriting alone runs 4-7% of gross proceeds, and skews to the high end for exactly the smaller deals that matter here (PwC) [8]. Add $3-5 million in legal, audit, and compliance costs, and a $100 million IPO costs roughly $10 million to execute [8], then $1-2 million every year thereafter just to stay compliant (Protiviti SOX survey) [9]. That cost structure mathematically excludes the median SME. It wasn't designed to include them. It was designed for the 5%.

And the 5% is exactly who gets served. By deal count, 95% of all M&A transactions are under $1 billion, yet mega-deals capture 57% of all deal value, and global M&A hit $4.8 trillion in 2025, the second-best year on record (Bain, 2025-2026) [10]. The liquidity exists. It just never reaches small companies.

Even the winners opt out

The companies that could IPO increasingly don't. The median company going public today is 12-14 years old, versus 4-5 at the height of the 1990s boom (Ritter) [7]. Median tech-IPO revenue is roughly ten times its 1999 level in real terms (Ritter) [7]. Stripe ($159B, whose founders call an IPO "a solution in search of a problem") and OpenAI ($852B) remain private as of this writing [11]. And SpaceX is the proof-point, not the exception: it created $1.7 trillion of value over 24 years before its June 2026 listing let the public own a single share (Renaissance Capital) [11]. By the time retail is allowed in, the growth has been harvested.

Retail wasn't invited anyway. Roughly 90% of IPO allocations go to institutions (industry data; even the SEC's Investor.gov concedes individuals struggle to get shares) [12]. And in private markets, the accredited-investor rules (thresholds unchanged since 1982) legally exclude 87.4% of individual Americans (SEC, June 2025) [12]. Those thresholds were never even inflation-indexed; adjusted, they would sit near $911,000 joint income and only 5.7% of households would qualify: today's "access" is definition drift, not policy [12]. Meanwhile private markets quintupled since 2010 to roughly $13 trillion (McKinsey, 2025) [13] while nearly nine in ten Americans were barred from touching them. The wealth was created. You just weren't allowed in the room.

The exit crisis nobody is pricing

Now the demographic detonator. McKinsey estimates 6 million American small and midsize businesses will change hands by 2035 (up to $5 trillion of value, 12 million jobs at stake) yet only 20-30% of businesses that go to market ever find a buyer (McKinsey Institute for Economic Mobility, Feb 2026) [13]. The largest US business-for-sale marketplace closed 9,586 transactions in all of 2025 (BizBuySell) [14]. Six million sellers. Nine and a half thousand closings a year. That is not a market. That is a queue for a door that fits one person at a time.

Stale photographs and settlement drag

Even for the companies inside the walls, the information system is archaic. A 10-Q arrives 40-45 days after quarter-end; a 10-K up to 90 days after year-end. Investors routinely price companies on numbers up to 135 days old. A 90-day-old photograph, delivered late [15]. Washington's response? The SEC proposed optional semiannual reporting in May 2026: less disclosure, not more [15]. Blockchain rails run the other direction: EY and Deloitte have both published on shifting assurance from retroactive snapshot to real-time verification [16]. Washington's answer is two reports a year. Ours is every block.

Settlement tells the same story. When the US moved from T+2 to T+1 in May 2024, NSCC's clearing fund fell $3 billion (23%) from shortening settlement by a single day (DTCC/SIFMA After Action Report, 2024) [17]. Atomic delivery-versus-payment settlement is the limit case: the BIS calls DvP on tokenized rails the canonical mechanism for eliminating principal risk between counterparties [18]. Let's be honest about the trade: fewer intermediaries means less counterparty and settlement risk, but self-custody moves responsibility to the holder. No chargebacks, no 1-800 number. We engineer around that with institutional custody options rather than pretending the trade-off doesn't exist.

Add it up. Hundreds of millions of companies with no market. A public market that halved. A toll booth priced for giants. An exit wave with no buyers. Disclosure getting slower while the technology for real-time attestation sits on the shelf. This is not a broken system that needs repair. It is a system working precisely as designed. For someone else.

The IPO isn't broken. It was just never built for 99.98% of the world's companies. We're rebuilding it as software.

Two Tokens, Two Visions

Most projects blur their tokens together and hope nobody asks hard questions. We do the opposite, because the delineation is the design.

The ICE and the NYSE

The cleanest analogy is Intercontinental Exchange and the NYSE. Owning ICE stock gives you the business that owns the exchange. A seat, a fee credit, a settlement rail. Those are working instruments inside the venue. $NPKN is the ICE share, and since Version 5.0 that is no longer an analogy: it is a common share of NewCo, the company that owns the machine, held and transferred on-chain. $RSX is the working instrument of the venue that machine is building. The Revenue Share Exchange, where revenue-share agreements become a listed asset class. One is what you own. The other is what the exchange runs on.

The dual-token lesson, learned honestly

Dual-token architectures have a graveyard and a hall of fame, and the difference is mechanical. VeChain's VET/VTHO works because the second token is consumption-driven: it is generated, spent, and burned by real usage, and it is not an emissions reward [19]. Terra's LUNA/UST failed because the two tokens were mint-burn pegged to each other. Each one's value was collateralized by faith in the other, and when faith moved, $40 billion evaporated in a week [20]. We name the anti-pattern because we designed against it: RSX will be consumption-driven like VTHO, never an emissions-reward token, and never (under any design we will ship) mint-burn pegged to NPKN. The two tokens share a builder, not a balance sheet.

The bridge: exactly what we can say, and nothing more

Here is the compliance-safe version, in full: Napkin intends to launch RSX. Any RSX distribution will be retroactive, snapshot-based, free, and its parameters announced only at snapshot. NPKN staking duration is expected to be an input to that snapshot. That is the complete statement. We do not publish allocation formulas, values, or yield figures for a token that does not exist, and we consider any project that does to be selling you something it cannot deliver (SEC/CFTC Release 33-11412, March 2026, says the same thing with more pages) [21]. The founder's RSX reserve is disclosed in advance as well. Section 11 has the target, the cap, and the structure. Transparent, not hidden.

The era roadmap

No price columns, no date promises on Era 5: RSX parameters are announced at snapshot. Not before. And the deal counts in the focus column are landmarks, not gates: an era does not wait for a counter.

Two clocks

The Book counts deals; the Eras count rails. An era turns when a capability ships (a sourcing stack, a shared infrastructure layer, an institutional rail), a gate of what the machine can do rather than how many deals it has closed. RSX is built from Era 1 on our own companies, the first customers of every rail we lay, and by the last era it is rails for rails: the venue itself becomes the product. Neither clock waits for the other.

The Napkin Machine

Strip away the token for a moment. What is actually here?

Think of Napkin as a playlist. Each balance sheet is a deal: a real company with real revenue, real EBITDA, and real bank accounts, acquired, integrated, and streamed into one consolidated portfolio. At the time of writing, 15 acquisitions have been completed since 2021, and the actively curated playlist holds a consolidated operating portfolio of 7-8 deals [3]. The token architecture maps the ambition directly: 10,000 deals at 100,000 tokenized shares per deal [22]. We are not there. We are a handful of deals in, and we will tell you exactly what stands between here and there.

The deal machine

Fifteen acquisitions completed since 2021 (from PV Labs in December 2021 to Creative Concrete & Excavating in April 2026) and here is the texture most acquirers hide: the portfolio has been actively pruned along the way. Several early digital acquisitions were divested back to their sellers for stock, sold, dissolved, or written off, leaving a consolidated operating portfolio of 7-8 companies at the time of writing: Levector, Skynet (Napkin India), EmberTribe, Podsicle, Jamm Media, Task Magic, and CCE [3]. More than a dozen completed transactions (including divestitures) is an honest scoreboard most acquirers never publish. We publish it because pruning is what disciplined capital allocation looks like from the inside.

The pipeline, at the time of writing, is concentrated where the machine now points: U.S. construction. CCE (Creative Concrete & Excavating) closed in Q2 2026 as the Day-1 platform of the Florida construction roll-up: a Tampa Bay grading and sitework operator with a 25+ year anchor relationship with Lennar Homes, roughly $28M of 2025 revenue and ~$4.5M EBITDA on the underwriting basis (~$28.8M trailing twelve months). GCUC (Gulf Coast Utility Contractors) is the combination target: a Northwest Florida underground-utilities contractor with ~200 employees, a 248-unit equipment fleet, adjusted 2025 revenue of ~$103.3M, adjusted EBITDA of ~$29.7M, and a ~$118M backlog: expected to close at the Q2/Q3 2026 boundary [3]. Rock Solid and Innerflow, two complementary operators, were referred to Napkin by CCE itself (the flywheel working) and are expected early Q3. Five further construction targets are in process, three with signed LOIs; USG and SSU remain pipeline targets. The current state of this list always lives in ongoing shareholder reporting, never in a static PDF.

Behind the pipeline sits the marketplace: NapkinDeals.com carries 16,268+ live listings representing roughly $54 billion in aggregate asking-price value across 50+ countries, with 110,744+ deals analyzed by the sourcing stack, alongside an operating history of 105+ clients served across the group [3]. Asking-price value is gross opportunity flow on the platform, not committed volume; we make that distinction ourselves before anyone else has to. The marketplace matters because in the sub-$50M market, deal flow is the moat. There is no Bloomberg terminal for a profitable sitework contractor in a secondary city. You find them, or you don't, and we have analyzed a hundred and ten thousand of them.

Deal structure is flexible by design. Consideration ranges from 25% to 75% cash, and the current average is roughly 50% cash, 25% equity, and 25% seller's notes: seller financing under which the seller effectively mortgages the business to Napkin, keeping skin in the game and reducing day-one cash requirements [3]. Napkin is amenable to 100% cash for strategic targets it likes, and equally to 100% equity. One mechanism is genuinely distinctive: the equity preference principle. Napkin prefers the equity component to be greater than zero, and gives preference, ranking, and onus to sellers who pledge their value with strong equity considerations. Those who show more faith rank higher, pro rata, in the acquisition queue. With tens of thousands of deals in the funnel, ranked algorithmically against the company's values, underwriting buy boxes, and strategies (the Gold Miner Thesis of Section 13, run through the AAA stack above) belief is priced into the deal algorithm. This is not financial engineering for its own sake. It is alignment. A seller who takes equity is not exiting; they are joining, with their net worth riding on the portfolio they just strengthened. In a market where, per McKinsey, only 20-30% of listed businesses ever sell [13], we show up as the buyer who closes and makes the seller a partner.

The construction pivot

The move from sub-scale digital agencies to asset-heavy, cash-generative U.S. construction is the platform's defining capital-allocation decision, and the logic compounds. Equipment-heavy operators carry tangible collateral, and collateral is bankable, exactly what makes the bond financeable and every subsequent acquisition cheaper to fund. The sector sits squarely inside the AI CapEx Supercycle: power, fiber, cooling, and water/sewer infrastructure for data-center build-out is grading, trenching, and underground-utilities work, precisely what the portfolio now does. And the targets are operator-led, succession-driven businesses: the exit wave of Section 2 made real, one retiring founder at a time. The screening funnel behind the pivot processed 60,000+ opportunities against a hard buy box (tangible assets, durable cash flow, LBO-serviceable economics) and produced four signed construction transactions [3]. That is the Gold Miner Thesis of Section 13 operating at institutional scale: the funnel monetizes what it rejects and buys only what survives.

The AI operating layer

The reason a platform this young can run at this cost structure: 62 AI employees across 8 departments (sourcing, diligence, legal prep, finance, integration, operations, reporting, support) built on our AAA framework (APIs, AI, Algorithms) [3]. The stack is real infrastructure, not slideware: NATS JetStream event streaming, Supabase, and Claude-powered bots doing the document-heavy work that mid-market M&A drowns in. The result is that the cost of evaluating deal number 110,745 approaches zero, which is the entire economic argument for buying small companies at scale. Investment banks can't touch a $3 million EBITDA business profitably. Software can.

The bond

The EUR 500 million senior secured asset-backed ETN program (signed December 25, 2025, subscription opened 15 December 2025, issue date 16 February 2026) is listed on the Vienna Stock Exchange as a private placement, with an initial series of roughly EUR 150 million. The terms, in full: an 8% fixed coupon (deferrable to maturity without event of default), a 7-year maturity to 15 February 2033 plus two optional 5-year extensions, ISIN XS3265939077, issuer PM Alpha DAC (Ireland), sub-arranger Plurimi Wealth (Monaco/Dubai), trustee Waystone, paying agent BNY Mellon (Dublin), and investment sub-advisor IDC (Delaware) [23]. The series carry first-ranking liens over operating assets, sponsor equity is fully subordinated, and eligible investors are professional and qualified non-U.S. purchasers under Reg S. Those first-ranking liens are exactly why the published waterfall and restricted-payments analysis of Sections 5 and 6 matter: the token's seniority position is defined against real secured creditors, in writing, before TGE.

One perimeter note, unchanged since Version 5.0: the Conversion of Section 6.7 moves the cap table, not the collateral. The construction book remains bond-collateralized inside the NewCo perimeter throughout and after the reorganization, and the bondholders' security is undisturbed by the change of ownership above it.

The capital arrives on a conveyor, not in one cliff. Subscriptions accumulate with the arrangers and auto-draw to Napkin in EUR 10 million tranches as each fills, tranche after tranche, as market distribution deepens [23]. Every filled tranche buys companies; every company adds attested cash flow; every dollar of cash flow feeds the Engine. The EUR 500 million program is the current envelope. The long-run ambition, as distribution matures across future series, is EUR 2 billion [3]. That figure is ambition, never current program size, and we will always label it that way.

The rails are multi-currency by design. The bond is issued in euros and deployed in dollars, where the bulk of the acquisitions close; international deals settle in whatever currency the seller values; and stablecoin settlement is welcome the day a seller prefers it. Currency is a settlement detail, not a strategy constraint.

The rails work for us, not us for the rails.

Here is the honest framing, because we would rather you hear it from us: the Vienna MTF is exchange-regulated, not a full regulated market, and a bond listing is not an audit anchor [24]. What the program gives us is real credit-market history: institutional trustee, paying-agent, and arranger relationships and a disclosure track record. What the token gives you is stronger: the NPKN terms carry contractual disclosure covenants that exceed the Vienna venue's requirements. Section 6.4 lists them, and they hurt us, not you, if we miss. We don't oversell the bond. We out-covenant it.

The Constellation question

Any serious reader will raise the comparison, so we will: Constellation Software spent three decades acquiring hundreds of vertical software companies and compounded famously by doing small deals, permanently, with discipline [25]. That is the genus we belong to. Now the attack: Constellation closed thousands of transactions over 30 years. You have 15, and you have already had to prune. Correct. Which is why our published intermediate milestone is 100 closed acquisitions, and why the Engine's economics step up on cumulative attested free cash flow, not deal count. You cannot game audited cash. At 15 deals we have proven the machine closes, and prunes. At 100 we will have proven it scales. Only then does 10,000 stop being a slogan and start being arithmetic. Judge us era by era.

The numbers, labeled honestly

Historical actuals are modest, and we publish them anyway: FY2025 consolidated revenue of $6.64 million CAD at a 50% gross margin, with the operating companies EBITDA-positive at +$0.58 million before holdco overhead (consolidated EBITDA -$0.51 million) [3]. Small, real, and (for the first time) configured to inflect.

Instead of a dated projection table, here is the run-rate staircase, labeled clearly as run-rate configurations rather than GAAP results:

• Stage 1: done. CCE closed: a ~$37.5M CAD revenue run-rate at ~$6.3M EBITDA. The first EBITDA-positive configuration in company history [3].
• Stage 2. In closing. GCUC plus the referred tuck-ins close: a ~$152.5M revenue / ~$38.1M EBITDA run-rate.
• Stage 3. The initial bond series deployed. Full deployment plus identified integration levers (USD 5.8-10.1M of EBITDA uplift) puts a $188M+ revenue run-rate platform on the board.
• The ambition. The $1 billion revenue platform and the 100-acquisition milestone, and then the long arc to 10,000.

At the time of writing the platform sits at roughly 15% of its initial $1B ambition, and for the first time the machinery to close the gap (capital, collateral, origination) exists. Stages 2 and 3 depend on financing and closing, and we publish every close, so you can score us in real time rather than take our word annually. Current figures always live in ongoing shareholder reporting and, post-launch, on the on-chain Proof of Balance Sheets dashboard.

The moat isn't the code. It's the balance sheets.

The cadence: the Book of 10,000 deals

The 10,000-deal ambition now carries a published cadence, and the cadence is a covenant. The Book closes in 50 chapters, each chapter 34 batches of seven deals; the seventh deal of every batch is a rest close, and every seventh chapter is a sabbath chapter of consolidation: no acquisitions, financing reduced, reserves distributed. After seven sevens of chapters, the fiftieth is the Release, and it opens with the Great Reconciliation: the published ledger of every error the machine made on the way, before any celebration. A machine that publishes its rest schedule is a machine you can hold to a work schedule.

The funnel that feeds the chapters is data, not meetings. Sellers enter as pledges, with pledged values visible on the dashboard before any capital moves, ranked by the equity preference principle described above. A thousand simultaneous pledges sounds like chaos; run through the funnel, it is roughly one working chapter of qualified closes. And every pledge gets an answer within 40 days: the 40-day answer covenant, scored publicly like everything else we publish.

Between pledge and close stand the Twelve Gates, the verification frame every deal walks through, and the evidence standard at each gate is two or three witnesses: no number enters the ledger on one source, and any number that drives the Engine requires three. The final gate is accretion: no batch closes unless modeled NAV per share after the close is at least NAV per share before it. The cadence sets the tempo; the gates make sure the tempo never outruns the truth.

$NPKN Tokenomics: The Engine

Tokenomics is where crypto projects hide the con. Low float, high FDV, cliff unlocks, yield printed from the token's own supply, buybacks announced and never executed. The market has read this playbook so many times that of 118 major token launches tracked in 2025, 84.7% broke below their TGE price, with a median drawdown of -71% (Memento Research, 2025) [26]. We designed $NPKN by studying every one of those corpses. This section is the autopsy report turned into a blueprint, and every mechanism in it is hard-coded, published, and attested before a single token trades.

One vocabulary note before the machinery. Because the token is a common share, everything in this section is corporate finance wearing its real name. Burns are share cancellations. Tender windows are issuer self-tenders. Staking distributions are opt-in dividends. The crypto mechanics and the corporate-law mechanics are the same mechanics, and that identity is the design.

5.1 Supply: One Billion Shares, Ten Thousand Deals

$NPKN has an authorized capital of 1,000,000,000 tokenized common shares, fixed in the token terms and the corporate constating documents alike. The number is not arbitrary. Our roadmap targets 10,000 acquired companies: 10,000 deals, 10,000 balance sheets consolidated into one audited perimeter. One billion shares divided by 10,000 deals is 100,000 tokenized shares per deal. Every closed deal adds a cash-flowing company to the playlist; the authorized count that owns the playlist never grows.

Legacy shares map onto this base through the Arrangement's exchange ratios, set under a fairness opinion and disclosed in the plan-of-arrangement materials before anyone votes. We will not publish a fake precise ratio in a marketing document; we publish the mechanism, and the court process publishes the number.

The share count is built to shrink. We publish a terminal supply target of 500,000,000 NPKN: a share-count reduction target, executed as buyback-and-cancel. Cancellations funded by the Engine (Section 5.4) reduce the outstanding count toward that number, permanently and traceably, until half the authorized base is retired.

Why publish a terminal number instead of a burn percentage? Because BNB proved that a countdown is the single most legible supply commitment in this industry. BNB launched at 200M supply with a published target of 100M, and has burned toward it with a transparent quarterly formula for years: down to ~136M by mid-2026 (BNB Foundation burn reports) [27]. Holders do not need to model fee curves. They can count. "Share count falls from 1B toward 500M, funded by attested portfolio cash flow" is a sentence a retail holder and a credit analyst can both verify, and unlike a crypto burn, a share cancellation is a corporate act with legal finality.

One honesty note before anyone else raises it: a burn detached from revenue is supply theater. Tron's "deflationary" design flipped net inflationary in Q1 2026 when usage shifted [28]. Our cancellations are funded exclusively by attested free cash flow: if the portfolio doesn't earn, the count doesn't shrink, and the dashboard will show exactly that.

One more covenant completes the supply design. The per-deal issuance budget is 50,000 tokens: the per-deal share of the issuable half of the cap, identical for every deal regardless of size. The largest platform acquisition and the smallest tuck-in draw the same budget from the register; no deal, however big or however early, gets a larger claim on the share count than any other. That is equal census treatment of the register. Pioneers are rewarded through the Accretion Ladder of Section 5.14, never by skewing the register.

5.2 Allocation: The End-State Cap Table, No Footnotes

The 2024-2025 launch data is unambiguous: supply overhang kills tokens faster than bad value accrual. The median 2024 launch floated just 12.3% of supply (Binance Research, 2024) [29]. Celestia's October 2024 cliff released 176M tokens and the price fell over 90% [30]. Ondo's annual mega-cliffs added 60% to circulating supply in a single day: twice [30]. We do the opposite of all of it.

Because the token is a share and the shares arrive in phases (the Arrangement's conversions, then treasury raises over years), the table below is labeled exactly what it is: an end-state target at full distribution, not a genesis print.

Say the caveat plainly rather than hiding it in a footnote: the Arrangement's exchange ratios and the sizing of future raises will move individual rows by a few points. Three covenants do not move, ever: the community holds the majority at full distribution, the founders are capped, and the Founder Liquidity Dial applies. Five rules govern the table, and each one is a lesson paid for by someone else:

The Founder Liquidity Dial is the spine. The founders roll the substantial majority of their holdings into tokenized shares in the Arrangement, and the founder side of the table is capped at 20% of supply, a cap that already contains its own future: every founder-level recruit the machine ever grants is paid from inside that pool, never from the community's side. Founder sales run on a published dial, not a discretionary window: up to 10% of remaining holdings per rolling 12 months, executed outside blackout windows, disclosed on-chain within days of execution, and founder participation in the Phase 1 tender counts against the same dial, disclosed in the token documents before TGE. The fairness is the size of the allocation, not a padlock. Prior cycles handed insiders 40 to 80% at genesis (XRP was roughly 80% insider; BNB gave 40% to the team plus 10% to angels), while the cohort of winning launches averaged 54% community. Our cohort model shows founders selling the maximum dial every single year and still ending near 9.5% of supply, with the community holding above 62%. Version 5.0 carried a hard Founder-Last schedule; Version 6.0 replaces the padlock with the cap, and publishes the arithmetic. "Everyone else gets value first; the founders take value last" survives as the capped-allocation ethos: the community's majority is written into the size of the slices, not enforced by a schedule.

Every vesting event lands on the 6th of the month. Monthly linear, fixed calendar date, never an annual cliff. Hyperliquid's team tokens vest ~1.2M per month, always on the 6th, and the market rewarded that predictability with a top-10 asset [31]. Unlock anxiety is a function of uncertainty. We remove the uncertainty.

Every purchaser in every raise gets identical terms. Same schedule, same calendar date, no better terms, ever. Berachain's disclosure that Brevan Howard held a private refund clause retail lacked destroyed that community's trust in a news cycle [32]. There are no side letters here. If we ever offer one, this document is the evidence against us.

The Treasury Escrow is mechanical, not discretionary. XRP's escrow (1B released monthly, the unused majority re-locked, all on-chain) is the proven structure for managing a large treasury position without a supply overhang narrative [33]. Our escrow releases on a published monthly cap; anything unspent re-locks automatically, visible on-chain.

Locked and unvested tokens cannot stake, cannot earn yield, and accrue zero RSX-snapshot weight. Founders included. This closes the hidden-emission bug that gutted Celestia, where locked investors could stake and sell the yield: Polychain extracted over $240M in staking rewards on nominally locked tokens (on-chain analyses, 2024-2025) [30]. A lock that pays out is not a lock. Ours don't pay.

5.3 Float and FDV: The Honest Paragraph

At TGE, at least 40% of the tokenized shares then outstanding are genuine, unlocked, tradeable float, distributed through the compliant lanes detailed in Section 9: EU prospectus retail, the BC offering memorandum, Reg D 506(c), and Reg S. No loaned-to-market-maker supply is counted in that number (Section 8.4).

And here is the sentence low-float projects never print: fully diluted valuation is real. Up to 1,000,000,000 authorized shares can eventually exist, and you should price the project on that basis, exactly as you would read a company's authorized capital and option pool. We state the FDV plainly at TGE and again in the risk factors, alongside the one discipline that governs all future issuance: treasury shares only ever issue above attested NAV (Section 5.6), so dilution is only ever accretive. The low-float/high-FDV structure (float a sliver, let the FDV headline do the marketing, unlock onto retail later) is now a recognized failure pattern with an 84.7% strike rate against buyers [26]. We will not run it. If our FDV looks expensive against attested portfolio cash flow, that is a fact you are entitled to see clearly, in our own document.

5.4 The Engine: The Waterfall That Cannot Be Renegotiated

Value accrual is where every governance token dies. Arbitrum passed 60+ proposals and routed nothing to holders: treasury rich, token at all-time lows [34]. Uniswap generated billions in fees for five years while UNI bled from $45 toward $3 waiting on a fee switch [35]. The market's verdict in 2026 is settled: accrual that depends on a future vote is worth approximately zero. So nothing in the Engine depends on a vote.

In corporate-law terms, the Engine is the published capital-return policy of the tokenized common equity, and it uses the three oldest tools in public-company capital allocation: buybacks that cancel shares (Section 5.5), issuer self-tenders at a NAV-referenced floor (Section 5.5), and cash distributions to shareholders who opt in via staking (Section 5.7). What makes it an Engine rather than a policy memo is that the tools are hard-coded, formula-driven, and non-discretionary.

The waterfall runs in one direction, published as one diagram, and executes in this order:

• Attested portfolio free cash flow enters the waterfall. Only auditor-attested figures, per Section 8. Self-reported numbers never touch the Engine.
• Bond obligations and maintenance come first. Interest and covenants on the EUR 500M Vienna-listed ETN program (whose series carry first-ranking liens over operating assets) plus maintenance capex, are paid before the token sees a euro. We say this explicitly: the token tranche is contractually junior to the bond. We publish the program's restricted-payments analysis (basket capacity, ratio tests, lien carve-outs, how the waterfall fits) and the Engine auto-suspends if bond coverage tests breach. A token flow that would breach a bond covenant is a default machine, not a value engine.
• The Token Cash Engine receives a fixed percentage of what remains: 20% at launch, stepping to 30% and then 40% as the portfolio hits cumulative attested-FCF milestones.

Note what the milestones are not: deal counts. "Unlocks per N acquisitions" is trivially gameable. A hundred shell purchases at one euro each "closes a hundred acquisitions." Our step-ups trigger on cumulative attested consolidated free cash flow. A hard-currency number, examined quarterly under ISAE 3000 [36], with auditor-verified "Qualifying" definitions (minimum trailing EBITDA, arm's-length pricing, at least two quarters inside the consolidation perimeter before anything counts). Cash is harder to fake than transactions. That is why we gate on cash.

The Engine is a covenant, not a promise. It cannot be voted up, voted down, or quietly turned off, and it pays bondholders first, in writing.

5.5 The 50/50 Split: Cancel One Half, Arbitrage the Other

The Token Cash Engine splits its market flow 50/50 between two mechanisms that attack the two distinct ways portfolio instruments die.

(a) Programmatic buyback-and-cancel. The per-share compounder. Because the token is a share, every burn is a share cancellation: a corporate act, legally real, permanently shrinking the outstanding count. The program is built to the EU market-abuse safe-harbor architecture, because we are already a MAR-regulated issuer (Section 8.5) and because discretion is the enemy: a pre-committed formula (a fixed share of last-attested FCF), executed by an independent agent, TWAP over the full period, daily volume caps, from a published wallet, with blackout windows around financial disclosures and acquisition announcements. Every purchase and every cancellation is traceable on-chain, reported monthly at trade level. Management never touches the button. A serial acquirer is a continuous generator of inside information, and discretionary buybacks on top of that are a manipulation machine unless procedurally neutered. Ours is neutered by design.

(b) Quarterly NAV self-tenders. The discount killer. Here is the uncomfortable truth about portfolio-backed instruments: without an exit channel at NAV, the discount is the equilibrium, not the exception. Listed private-equity trusts (a century-old TradFi control group for exactly what $NPKN is) trade at persistent 20-40% discounts to NAV; HarbourVest sits near -30% and announced a $500M distribution plan specifically to fight it (Jefferies; LSE trust data, 2025-2026) [37]. By March 2026, roughly 40% of publicly traded Bitcoin treasury companies were below NAV (NYDIG, 2026) [38]. BlackRock's BUIDL never trades at a material discount for one reason: a 24/7 redemption channel at NAV that authorized participants arbitrage instantly [39].

We cannot offer instant full redemption against illiquid SMEs: anyone who promises that is selling you a duration mismatch. What we can build is the ETF-style loop at a cadence the assets support: every quarter, Engine funds open a Dutch-auction issuer self-tender, executed at the better of market price or a discount-bounded attested NAV, capped per quarter and priced only against last-attested figures with a notice period (so nobody farms stale marks). If $NPKN trades at a deep discount to attested NAV, the tender is a funded, recurring, mechanical bid that scales with the discount. Arbitrageurs do the rest. That is not price support. It is the closed-end-fund self-tender and the ETF creation/redemption logic in one loop, adapted to quarterly attestation reality, run by the issuer whose shares these are.

5.6 NAV Discipline: The Rule Strategy Learned Too Late

Codified in the token terms and the corporate policy alike: no new share issuance of any kind (M&A currency, treasury sales, anything) while the token trades below last-attested NAV per share. Issuance above NAV is permitted only to fund accretive acquisitions.

Strategy (MSTR) ran the issue-above-NAV flywheel brilliantly until belief broke; by June 2026 it traded at roughly 0.72x enterprise mNAV, where every share issued destroys holder value and the flywheel runs in reverse (NYDIG, 2026) [38]. Metaplanet had to bolt on a buyback rule after a 96% mNAV decline [40]. We install the discipline before the discount exists, not after, and we point the flywheel at operating companies instead of a treasury asset: Strategy's per-share compounding logic, run on real businesses with attested cash flow. Below NAV, we buy. Above NAV, we may issue, but only into deals that grow NAV per share. There is no third mode.

Two sentences of valuation doctrine before anyone imports the wrong framework: NAV is the ratchet, not the appraisal, and tokenized common equity of an operating consolidator is priced by the market on earnings like any public company, so any earnings arithmetic we publish is conditional sensitivity, never a target. Treasury-company mNAV framing is explicitly the wrong lens here: mNAV prices a wrapper around a treasury asset, while this is an operating business whose shares the market will price on cash flow.

The M&A settlement utility follows the same discipline, and Version 5.0 collapsed a distinction: the equity component of acquisition consideration (typically around 25% of a deal, and ranging up to 100%) is settled in tokenized NewCo shares, because Napkin equity and $NPKN are now the same instrument [3]. We promise no fixed percentage of aggregate purchase price. But because the deal-ranking algorithm of Section 4 rewards sellers who elect larger equity components, structural demand scales with seller conviction: the more the sellers believe, the more of every deal settles in shares rather than cash. Seller notes reduce day-one cash needs further, complementing the bond, and any share issued into a deal obeys the rule above: above attested NAV only, accretive only. Section 5.12 shows what this becomes at scale.

5.7 sNPKN: Real Yield or No Yield

Staked NPKN (sNPKN) is a standard ERC-4626 vault [41]. Yield is paid in stablecoins, sourced from the Token Cash Engine's distribution leg: actual portfolio cash flow, converted and distributed. In corporate-law terms these are dividends: cash distributions to shareholders, made opt-in through staking so that holders choose between income now and compounding through cancellations. The yield is never, under any circumstances, paid in emitted NPKN.

This is a hard rule with a specific tombstone on it. GMX invented the "real yield" narrative in 2022, then ran parallel esGMX emissions that pushed effective inflation above 10% a year. Holders eventually did the dilution math; the token fell 81% from its high [42]. Yield paid in an escrowed version of your own token is recycled dilution wearing a yield costume. If the Engine's stablecoin flow is modest in early quarters, sNPKN yield will be modest, and the dashboard will show the honest number. We would rather publish a small true yield than a large fake one.

sNPKN is also liquid. Time-weighted multipliers reward longer holding, but there are no multi-year hard locks. Curve's 4-year veCRV locks built a governance economy and trapped retail inside it; Pendle publicly abandoned ve-locking for liquid staking in January 2026, and the entire industry followed [42]. Loyalty should be rewarded, not imprisoned.

Alongside sNPKN sits a second, fully optional lane: term-locked Coupon Vaults at 6, 12, and 24 months. The vaults accept stablecoin deposits that co-fund acquisitions alongside the bond: credit extended to the machine by its own shareholder base, on disclosed terms, targeting bond-like stablecoin yield under the same hard funding rule as sNPKN: paid only from Engine distributions and, as bond series retire, from the freed coupon (Section 6.2 has the retirement arithmetic). Variable, funded by actuals, never promised. And income is opt-in throughout the design: stake for distributions, or hold unstaked and let the cancellations compound your ownership of the machine instead. Both are legitimate ways to hold the same share.

5.8 Not a Ponzi: Sources and Uses, Segregated and Attested

The first test any quant runs on a "token backed by assets" is circularity: does token-sale money buy the assets that back the token, or pay for earlier holders' exits? If yes, the collateral is the buyers' own money. A Ponzi with an accounting metaphor. We fail that test on purpose, in the right direction, and the Conversion makes the answer structural rather than rhetorical.

The iron rule, published as sources-and-uses and attested by the auditor: exit cash comes from professional fiat capital that knowingly underwrote it; token proceeds only ever buy new companies. Unpack it into its three legs. First, every legacy shareholder exit in Phase 1 of the Conversion is funded by fiat: the new NewCo equity raise, or the bond program's earmarked buyback allocation, never token-sale proceeds, so no token buyer's money ever pays a prior holder to leave. Second, the existing portfolio arrived inside NewCo by court-approved share-for-share conversion in Phase 2, not by purchase, so the asset base backing the token at TGE was never bought with token money. Third, every tokenized-share raise at and after TGE deploys into new acquisitions, liquidity provisioning, technology, and Engine reserves under the NAV discipline of Section 5.6: issuance above last-attested NAV only, into deals accretive to every existing holder, so new capital strengthens per-share backing instead of circularizing it. A holder's money is never the collateral behind their own claim, and never another holder's exit. The machine predates the token, which is precisely the point.

5.9 What Buybacks Are Not

Read this paragraph twice, because we will never contradict it in any marketing material: accrual mechanisms do not defend price. Uniswap burned $596M of UNI in December 2025 and the token hit an all-time cycle low of $2.90 two months later [35]. Pump.fun spent 100% of revenue on buybacks and the price fell anyway [43]. Hyperliquid (the best-executed buyback program in crypto history) drew down through the H1 2026 bear despite ~$65M per month of automated buying [31]. Buybacks cannot outrun market beta and they cannot outrun deteriorating fundamentals, and anyone who tells you otherwise is lying to you or to themselves.

What the Engine actually does is compound per-share economics: every cancellation permanently increases each remaining share's ownership of a growing portfolio's attested cash flow. Not "the same logic as" an equity buyback at a well-run company: it is an equity buyback at what we intend to prove is a well-run company. That compounding is real, measurable, and published quarterly as NAV per share and FCF per share. Price is the market's job. Per-share cash flow is ours.

5.10 Design Choices vs. the Graveyard

5.11 The Capital Stack: Four Engines, One Machine

Zoom out from the token for one subsection, because the token is one lane of a four-lane capital machine, and holders deserve to see all four at once.

Debt: the bond conveyor. The EUR 500M Vienna-listed program of Section 4, drawn in EUR 10M tranches as subscriptions fill [23]. Senior, secured, disclosed: first-ranking liens, a published waterfall, and the restricted-payments analysis of Section 5.4. Bond capital buys companies; the token never ranks ahead of it and never pretends to.

Equity: the rolled base plus the NewCo raise. Roughly $19.8M CAD of outside capital invested by roughly 78 shareholders built the base [3]; the Conversion resolves that table into a single class of common. The self-selected holders who roll become tokenized shareholders in the Arrangement, and a new fiat equity raise into NewCo (targeted at roughly $50M) funds every Phase 1 exit and the legacy liabilities while leaving growth capital on the balance sheet, with zero added leverage. The pitch to that raise is pre-IPO equity where the IPO is on-chain and milestone-gated.

Tokenized-share raises: TGE and after, above NAV only. From TGE onward, new shares issue from treasury through the compliant lanes, under one gate: issuance above last-attested NAV per share, into uses that are accretive to every existing holder, per the NAV discipline rule of Section 5.6 and the sources-and-uses rule of Section 5.8. Below NAV, treasury never issues. Above NAV, every raised dollar buys new companies.

Staking: the community's fourth lane. Alongside sNPKN, optional term-locked Coupon Vaults (Section 5.7) accept stablecoin deposits that co-fund acquisitions alongside the bond: community credit on disclosed terms, targeting bond-like stablecoin yield, variable, funded by actuals, never promised. One hard guard governs the lane: NPKN itself is never borrowed against. No reflexive leverage, ever. A token collateralizing debt against its own machine is how $40 billion evaporated in a week [20], and that pattern is designed out of this stack at the root.

Notice what Version 5.0 did to the middle two lanes: the equity engine and the token engine are now the same instrument at different stages. Private common shares before the Arrangement, tokenized common shares after it, one class, one register, one NAV. Four engines, one machine, one waterfall, and one set of attested numbers that all four report against.

5.12 The Acquisition Currency: The Answer to 10,000

Serious readers ask the same question in the same tone: how does anyone get from 15 acquisitions to 10,000? This subsection is the answer, and it is the reason the token exists at all.

Sellers already take Napkin paper. The deal record of Section 4 proves the demand: consideration averages roughly 25% equity, sellers who pledge more equity rank higher in the queue, and some deals close at 100% equity [3]. But private stock is hard consideration to accept: illiquid, hard to value, impossible to sell down, dependent on an exit that may never come. A tokenized common share with a quarterly attested NAV, a funded tender floor, and live regulated venues is strictly better consideration than the private stock those sellers already say yes to. Version 5.0 did not create the acquisition currency. It upgraded the one that has been closing deals since 2021.

The upgrade turns consideration into a flywheel:

• Raise above attested NAV, per the discipline of Section 5.6.
• Buy companies; their attested cash flow joins the perimeter.
• The Engine buys back and cancels shares; NAV per share compounds.
• Compounding builds credibility; credibility makes the paper better currency.
• Sellers accept more token and less cash per deal; every raised dollar closes more deals.
• Repeat, faster each turn.

Constellation Software closed thousands of acquisitions with legendary discipline, paying overwhelmingly in cash, and it took roughly three decades precisely because cash consideration scales linearly with the balance sheet [25]. Liquid equity consideration scales with belief in the machine, verified quarterly by attestation, and it is the only currency that scales to 10,000 balance sheets. There is a compounding social effect too: every acquisition settled in token mints new believer-holders, because under the equity-preference principle the sellers who take the most token rank highest, so the shareholder register steadily fills with the operators of the very companies the machine owns. The 100,000 tokenized shares per deal of Section 5.1 stop being a metaphor. They are the deal currency, one closed deal at a time.

5.13 Soil and Harvest: The Operating Covenants

The Engine governs what the machine pays out. A second set of covenants governs how the machine treats what it owns, because a consolidator that strip-mines its own portfolio is a liquidation with better branding. Five rules, all published, all reported against:

The reinvestment share. One seventh of distributable free cash flow stays inside the operating company that earned it: maintenance beyond the minimum, product, people, incremental rotation of the asset base. The platform harvests the field; it never harvests the soil.

The reserve contribution. 10% of distributable FCF flows into a contingency reserve. The reserve is drawn on shortfalls, before any new credit facility is opened, and on sabbath chapters. Stored surplus is cheaper than emergency credit, and the dashboard shows the balance.

Rest closes take nothing off the top. In a rest close, the platform charges no management fee and pays no bonus. Rest is rest for everyone, including the people who run the machine.

Sabbath chapters retire financing. Each sabbath chapter retires one seventh of outstanding financing. The machine's leverage has a scheduled ebb built into its rhythm, not just a covenant ceiling.

The Release settles at par. At the Release, outstanding earn-outs settle at 100% with their conditions waived: the platform releasing the claims it holds over sellers, not the reverse. A machine built on sellers' trust periodically proves the trust runs both ways.

5.14 The Funding Truth

Ask the question every serious reader should ask: who actually pays for 10,000 companies? The answer that kills most token projects is "the token." Ours is arithmetic, and the token was never the wallet.

Start with the deal structure. Sellers finance roughly 30% of their own exits, through seller notes and deferred earn-outs: standard consideration in this market, and the strongest possible signal of seller conviction. Next, every acquired company collateralizes its own purchase. Companies bought near 3.6x EBITDA support secured borrowing near 2.5x EBITDA on day one: classic buyout mechanics, with the consequence that each closed deal expands the base that buys the next one. Recycled free cash flow, net of the covenants of Section 5.13, covers most of the rest.

What about the token? Run the refusal arithmetic. Funding the journey's cash bill from the issuable 500M shares would require an average of roughly $199 per token: about 12x the journey's median NAV. A plan that needs its own token to trade at 12x asset value in order to function is not a funding plan; it is speculation wearing a funding costume. Worse, aggressive early selling raises less money, not more: the fixed cap forces cheap early sales that exhaust supply before it is worth anything.

So the treasury runs the opposite policy, published as the Accretion Ladder: the per-deal sale budget of Section 5.1 stays at its base level in ordinary markets, triples above 10x attested NAV, and runs at five times base above 25x attested NAV. The ladder monetizes mania, never chases warmth, and never sells fear. The stress case is designed in rather than assumed away: in the permanent-bear scenario the machine completes with zero tokens sold, because the deal math above never needed them.

The secured layer completes the picture, and it is designed to be issued, not borrowed. The financing each acquisition supports is structured as asset-backed profit-sharing notes sold to the community in tranches: the community replaces the external lenders and collects the profit share itself, with a published target of 75% community-held. Same balance sheets, same security packages, a different creditor: the people who already own the equity.

What You Actually Own: The Legal Claim

Here is the question most white papers spend forty pages avoiding: what do I legally own, against which entity, and what happens if that entity fails?

The industry's usual answer arrived in July 2025, when Robinhood distributed "OpenAI tokens" to EU retail. OpenAI publicly disavowed them within days. They conferred no equity, no redemption right, no claim on anything [44]. Holders of that product own exposure to a story. When the story ends, they own nothing. Every token that describes its backing with metaphors instead of contracts is running some version of that trade.

We are not. Here is the claim chain, and it will be published as constating documents and binding token terms, not just as a white paper diagram.

6.1 The Instrument

$NPKN is a tokenized common share of Napkin 2.0 ("NewCo"): common equity of the company that owns the machine, in digital form. Not a derivative that references a share. Not a participation note that mimics one. Not a wrapper around somebody else's stock. The share itself: entered in NewCo's share capital, maintained on a register kept by a registered transfer agent, wrapped in the ERC-3643 permissioned standard of Section 7, and carrying the same rights as any common share, because the holder is not "like" a shareholder. The holder is a shareholder. The chain of claim runs:

• You hold the token. The token is the digital form of a registered common share of NewCo. The share register and the chain move together, token for share, one to one, with the compliance layer of Section 7 enforcing who may hold it.
• NewCo owns Napkin. Through the Conversion of Section 6.7, NewCo holds Napkin Inc. and its subsidiaries: the portfolio, the marketplace, the AI stack, the whole consolidation perimeter of Section 8.

Two links. Version 4.2 needed three links and a participation contract in the middle; Version 5.0 deleted the middle. There is no contractual claim standing between you and the machine, because you own the machine.

Say the flip in plain language, because four versions of this paper said the opposite with equal plainness. Earlier versions drew a careful line: token holders own the machine's output, not the machine's title. That line is now erased. Token holders own the machine. The Engine of Section 5 is not a payment you receive from an owner; it is your company's published capital-return policy, executed on your shares. And this is not a legal frontier: SEC-qualified tokenized common shares have traded with a registered transfer agent since 2021 [107], and the 2025-26 tokenized-stock wave industrialized the distribution rails [108]. Most of that wave wraps other issuers' shares in exposure products, which is exactly the Robinhood lesson above. $NPKN is the opposite construction: issued by the company whose shares they are, with the issuer's own obligations attached.

One senior claim is disclosed rather than discovered: the series of the EUR 500M bond program hold first-ranking liens over the operating assets of the construction book, inside the SPV perimeter the program finances [23]. Bondholders hold that security; shareholders hold title; the Conversion leaves the security undisturbed. Section 6.2 states exactly what the ranking means if things go wrong.

6.2 Insolvency, Stated Plainly

In an insolvency of the group, NPKN holders hold common equity, ranked as equity: behind the EUR 500M Vienna-listed ETN program's secured series and their first-ranking liens [23], behind every operating liability of the perimeter, last in the waterfall. That is the whole sentence, and it is what ownership means. Creditors are paid in full before shareholders receive anything; whatever remains belongs entirely to the shareholders. Last claim, full residual. We will not dress the rank up, and we will not apologize for it either: equity's position is the price of equity's upside, and since Version 5.0 this paper states the price in the same breath as the title.

We state this bluntly because the alternative (silence) is what converts holders into surprised victims. Equity honestly ranked and honestly disclosed is an investable instrument. An undefined claim is a lottery ticket with extra steps.

The stack above you is also designed to shrink. The balance sheet is built to retire its own debt: every bond euro repaid frees 8% in annual interest for the Engine and loosens the covenants above the equity, so the debt overhead is a starting position, not a life sentence. And future bond series can be issued as tokenized notes with priority access for long-term stakers (eligibility rules apply), so that over time the community replaces the external lenders and collects the interest itself. One honest caveat travels with that privilege: a holder who owns both the share and a tokenized bond has stacked correlated risk on one machine. That is why it is an opt-in privilege, stated plainly, never a default.

6.3 Why We Embrace the Security Label

Most crypto projects structure around securities law. We structure into it, and Version 5.0 retired the last ambiguity: a common share is not "arguably" a security, and it is not a security "by design". It is the paradigm case, the instrument securities law was written about. We spent four versions engineering an instrument honest enough to deserve the label. Now we simply hold the thing the label names.

The regulatory routing survives the recast intact, which is the quiet advantage of tokenized equity: a tokenized share is a financial instrument under MiFID II, and MiCA's Article 2(4)(a) explicitly excludes such instruments from its scope, which routes $NPKN into the EU Prospectus Regulation [46]. That sounds like a burden. It is actually the unlock: an approved prospectus, passported across 30 EEA states, is the only legal path to marketing an instrument like this to European retail. The "utility token" costume (economic rights in substance, governance label on the tin) is precisely what dies in enforcement. BaFin's 2025 shutdown of Ethena's EU offer is the template [47], and Section 9.4 walks through it. Ethena launched, then negotiated. We pre-file, then launch.

There is a deeper point. Our entire pitch is that you own the machine. Corporate law and securities law are the machinery civilization built to make exactly that ownership enforceable: registers, fiduciary duties, disclosure, dissent and oppression remedies, a court that will hear you. Dodging that machinery would mean telling you the ownership story is true while structuring it to be false. The security classification is not our compliance cost. It is our product working as described.

6.4 The Disclosure Covenant Stack

The Vienna bond gives us capital-markets discipline, but we will not oversell it: the Vienna MTF is exchange-regulated, and its ongoing disclosure requirements are light. So the token terms carry contractual disclosure covenants that exceed any venue requirement:

• Annual audited IFRS consolidated financial statements from a named top-10 audit firm, delivered within 120 days of year-end.
• Semi-annual reviewed interims and quarterly ISAE 3000 attestations of the metrics that drive the Engine (Section 8) [36].
• Penalty inversion: if we miss the 120-day covenant, the Engine percentage automatically steps UP, with a mandatory self-tender trigger for sustained failure. Every disclosure regime we studied punishes holders for issuer opacity. The numbers go dark and the price pays. Ours is inverted by contract: disclosure failure costs the issuer cash flow, immediately and mechanically.

If we go quiet, we pay more. Disclosure failure is a cost we imposed on ourselves, in writing.

6.5 The Issuer: A British Columbia Corporation

NewCo is incorporated in British Columbia, Canada, and the domicile is tax logic, not flag preference. The Conversion's second phase is a plan of arrangement under BC corporate law (Section 6.7), and a share-for-share exchange into a Canadian corporation preserves the rollovers embedded in the legacy table: advisors who earned shares over years, and sellers who took equity under s.85 deferrals, roll into NewCo without crystallizing the gains a direct token-for-share swap would trigger. That is the entire reason the structure is share-for-share first and tokenization of the share second; the order is the tax planning. The BC choice also compounds with everything in this paper that already lives there: the arrangement court, the founder's residency, and the uncapped BC offering-memorandum lane of Section 9.

EU retail distribution is stated as intent plus options, because that is what it currently is. The intent: a passported EU Prospectus Regulation prospectus using the existing EEA architecture of Section 9, with FMA Liechtenstein as approving authority and the Zug and Vaduz entities retained in the stack for exactly this job. The options: a direct passported prospectus for the tokenized BC share, or an EEA depositary instrument issued by Napkin Securities AG wrapping the share one-to-one, with TVTG making that token the legal carrier of the wrapped right [45]. Which wrapper ships is a decision for securities counsel inside the prospectus process, and the answer will be published in the token documents. We present the fork honestly rather than manufacture certainty we do not yet have.

6.6 The Critical Path, In One Table

None of this executes itself, and we would rather list the dependencies than let a diligence team discover them:

Any of these can slip, and if one slips, the token slips with it. That is not a caveat we mumble. It is Gate G1 of Section 10 working as designed: dates slip before gates do.

6.7 The Conversion

At the time of writing the claim chain of Section 6.1 does not exist yet: Napkin Inc. carries a conventional private cap table built over years of equity raises, advisor grants, and seller rolls. The Conversion is how that table becomes tokenized common equity without a single holder being wronged, and it runs in three phases, in strict order, each one a precondition for the next.

Phase 1: The Trim

A voluntary, fiat, pre-token tender that has nothing to do with crypto. Every legacy shareholder tranche is offered cash at or above invested cash basis, priced tranche by tranche (option and low-basis positions lowest, high-basis positions highest) under the company's confidential buyback model, which prices a full clearing of the legacy table at roughly $21.7M [3]. The holders who carried the early risk and want cash are honored with a cash win before the token era begins. Their exit is the last act of the old covenant, kept.

The funding rule is the iron rule of Section 5.8, applied at the source: exit cash comes from professional fiat capital that knowingly underwrote it; token proceeds only ever buy new companies. The primary source is a new equity raise into NewCo, targeted at roughly $50M: it clears every exit (roughly $21.7M), retires the legacy liabilities (roughly $8M), leaves roughly $20M of growth capital on the balance sheet, and adds zero leverage [3]. The pitch to that capital is pre-IPO equity where the IPO is on-chain and milestone-gated. The fallback is the bond program's earmarked $25M buyback allocation [23]. Token-sale proceeds are never a source, under any scenario, in any phase.

The Trim runs under a covenant with a name, published in advance, because unnamed principles get quietly shaved:

The Nobody Loses Covenant.

• Cash Floor. No shareholder is ever cashed out below hard cash invested per share.
• Universal Roll. Every holder may exchange paper for paper at ratios that preserve proportional ownership. Upside is never confiscated.
• No Forced Cash. Tag-along and drag-along mechanics only ever drag a holder into the roll, never into cash below basis.

Behind the covenant stands the Founder's Backstop: any holder whose contractual exit falls below invested basis is made whole personally by the founder. Section 13 carries both as stewardship in action; here they are terms.

Phase 2: The Arrangement

Paper for paper, no cash. A court-approved plan of arrangement under BC law converts the remaining, self-selected cap table into tokenized common shares of NewCo, collapsing every legacy class and instrument into a single class of common in one step. The tax logic dictates the shape (Section 6.5): share-for-share into a Canadian NewCo preserves the rollovers; a direct token-for-share swap would crystallize them; therefore the share converts first and tokenizes second. The reorganization exemption covers the conversion for existing holders, the fairness opinion sets the exchange ratios, and the court order binds them, with dissent rights protecting anyone who objects (Section 9.1).

The founders intend to roll the substantial majority of their 30M combined shares [3]. Founder flexibility before TGE is a published dial, not a hidden choice: the founders may sell into the Phase 1 tender on identical tranche terms up to a stated cap that counts against the Liquidity Dial's first rolling year (target: no more than 10% of holdings), the balance rolls, and the final split is disclosed in the token documents before TGE. From TGE onward, the Founder Liquidity Dial (Section 5.2) takes over: founders are capped at 20% of supply, with future founder recruits granted from inside that pool, never from the community; sales run on a published dial of up to 10% of remaining holdings per rolling 12 months, executed outside blackout windows and disclosed on-chain within days. The fairness is the size of the allocation, not a padlock. Prior cycles handed insiders 40 to 80% at genesis (XRP roughly 80% insider, BNB 40% team plus 10% angels) while the winning-launch cohort averaged 54% community, and the cohort model shows founders selling the maximum dial every single year still end near 9.5% of supply while the community holds above 62%. Everyone else gets value first because everyone else got most of the supply. Mechanism, not promise.

The perimeter: NewCo acquires Napkin Inc. and its subsidiaries, subject to bondholder change-of-control consent on the EUR 500M program, the critical-path item of Section 6.6, stated honestly rather than buried. The construction book remains bond-collateralized inside the perimeter, and the bondholders' security is undisturbed by the whole sequence.

Phase 3: The Raise

TGE. New tokenized shares issue from treasury through the compliant lanes (the passported EU prospectus, the BC offering memorandum, Reg D 506(c), Reg S) to fund the scale-up, and every raised dollar buys new companies. There is no circularity to find, because the sequence removed it structurally: the exits were fiat-funded in Phase 1, and the legacy book arrived inside NewCo by conversion in Phase 2, not by purchase. At TGE the token is backed by real equity in an operating company with attested revenue from day one. Product before token, and real backing: both satisfied as facts, not aspirations.

The pledge sequencing of earlier versions survives inside these phases, because the principle never depended on the instrument. Before anything is on-chain, the covenant, the arrangement documents, and the constating documents are the pledge: the contract IS the pledge. Before anyone can buy, Proof of Balance Sheets goes live and the attestations are hashed on-chain: numbers before money. At TGE the Engine contracts switch on and the pledge becomes self-executing: attested cash flow in; buybacks, cancellations, and tenders out; no discretionary hand on the valve.

The blockchain does not create the obligation. It makes the obligation impossible to hide.

Technical Architecture

Most token papers treat the chain section as a branding exercise. We treat it as an engineering decision with a paper trail. Every choice below follows the same rule: go where the liquidity, the custody, and the auditors already are, and say out loud where they are not.

Chain selection follows liquidity and custody, not narrative.

7.1 Hub and Spokes

$NPKN is one canonical asset with one canonical supply, issued once, on Ethereum L1. Everything else is a spoke: a burn-and-mint representation whose supply nets to zero against the hub. No wrapped tokens, no lock-and-mint bridges holding honeypots, no ambiguity about where the real thing lives.

The hub: Ethereum L1, ERC-3643

Canonical $NPKN is an ERC-3643 (T-REX protocol) permissioned security token that preserves the full ERC-20 interface. ERC-3643 is not a boutique choice. It is the consolidated institutional standard for compliance-embedded tokens: $32B+ tokenized across 200+ deployments and 180+ jurisdictions, with Tokeny as reference implementer, DTCC ComposerX integration, and the SEC chair citing the standard by name in July 2025 [48]. The BaFin-licensed 21X exchange in Frankfurt (where we intend $NPKN to trade) settles ERC-3643 assets natively [49]. Every transfer runs through an on-chain eligibility check against an identity registry: KYC claims, jurisdiction rules, investor caps, enforced in the token itself rather than promised in a PDF.

Why Ethereum L1 and not somewhere cheaper? Because canonical issuance is not a cost decision, it is a credibility decision. Ethereum still hosts more than 56% of tokenized RWA value [50], and every institutional issuer that matters (BUIDL, Ondo, Backed, Centrifuge) issues canonically there. Custody is first-class: Fireblocks, BitGo, and Anchorage support ERC-20/3643 assets out of the box, and tokenized-fund AUM under qualified custody crossed $20B in early 2026 [50]. A token that wants to be held by regulated intermediaries goes where regulated intermediaries already hold things.

The retail spoke: Base

Retail users should not pay L1 gas to hold a security token. The Base spoke uses burn-and-mint via ERC-7802 hooks: the crosschain mint/burn interface that ships in OpenZeppelin v5.5+ as ERC20Bridgeable [41], with Chainlink CCIP as the transport. The design point matters: ERC-7802 lives in the token, the bridge vendor does not. No LayerZero OFT.sol or any other vendor code is inherited into the token contract; CCIP is an authorized caller that can be replaced without touching $NPKN itself. The market has already voted on this pattern: $7.2B+ in assets migrated from LayerZero to CCIP since May 2026 [51]. Base gives us sub-cent fees and the fastest-growing RWA TVL of any network over the past 18 months [50]; we note honestly that its sequencer is centralized, which is acceptable for a distribution spoke and would not be acceptable for the hub.

The institutional spoke: XRPL mainnet, native

The third leg is a native issuance on the XRP Ledger mainnet, and we want to be precise about what that means, because the XRPL story is really two stories.

Story one is genuinely impressive. XRPL mainnet tokenized RWA value grew from under $1B in early 2026 to roughly $3.5B by July 2026, leading all chains on 90-day RWA inflows at +$1.9B [50], built entirely on native protocol primitives, not smart contracts. Multi-Purpose Tokens (XLS-33, live October 2025) are a purpose-built institutional standard with on-chain metadata, transfer restrictions, and KYC-authorized-holder-only issuance; Credentials (XLS-70) put KYC authorization on-ledger; Permissioned Domains (XLS-80, activated February 2026) enable gated venues; authorized trust lines and clawback have been live for years [33]. Ondo's OUSG mints and redeems there via RLUSD; OpenEden runs tokenized Treasuries; Guggenheim issues digital commercial paper; Archax, the FCA-regulated exchange, committed to bringing $1B+ of tokenized assets onto the ledger [33]. In May 2026, Ondo, JPMorgan's Kinexys, Mastercard, and Ripple executed the first cross-border, cross-institution redemption of a tokenized Treasury fund in under five seconds [33]. Around the ledger sits real institutional plumbing: BNY custodies RLUSD reserves (the fastest-growing regulated stablecoin), the OCC conditionally chartered Ripple National Trust Bank in December 2025 [52], and Ripple Prime clears over $3T annually for 300+ institutional clients [33]. That is a settlement environment worth plugging into, and $NPKN's XRPL spoke is a native MPT issuance inside a Permissioned Domain, with RLUSD as the mint/redeem and settlement rail.

Story two is the one most papers would omit. The XRPL EVM sidechain (mainnet since June 30, 2025) holds $25,741 of TVL as of July 14, 2026 [28]. Not million. Thousand. Zero DEX volume, ~168 active developers against Ethereum's ~8,448 [28], flagship launch dApps at zero TVL, a roughly 24,000-to-1 miss against its own ecosystem projections. We do not deploy there, and we say so in print because the reasoning generalizes: institutional demand found XRPL through its native primitives and skipped the EVM layer entirely. A chain decision made for narrative alignment rather than measured liquidity is how treasuries get stranded. Accordingly, the XRPL spoke activates against three published, checkable gates: the XLS-66 native lending protocol going live (in validator voting as of July 2026) [33], MPT custody support from a top-tier independent qualified custodian beyond Ripple's own stack, and Archax-scale secondary liquidity actually materializing. Until then it is built, tested, and waiting. Not marketed.

7.2 Contract Stack

The stack is OpenZeppelin Contracts v5 (v5.6.x line), Solidity ^0.8.26, built and tested in Foundry. The architectural rule: immutable core, modular periphery. The token contract itself is not upgradeable: holders should never have to trust that an admin key will not rewrite the asset they hold. Everything that legitimately needs to evolve lives in swappable modules behind interfaces: the compliance module (jurisdiction rules change), the Engine module (milestone steps activate), and the oracle adapter (feeds get upgraded). Module replacement is gated by a Safe multisig behind a 48-hour timelock, so every change is publicly visible for two days before it can execute. Emergency pause is the one exception: held by a security-council Safe with no timelock, because an incident does not wait 48 hours.

The Engine, as code

Section 5 described the Token Cash Engine in economic terms. Here is the same machine as a call flow:

• The oracle adapter receives an auditor-attested NAV and consolidated-FCF figure from the Chainlink feed and calls attestNAV() on the Engine.
• Stablecoins arrive at the Engine from the distribution waterfall. The Engine's share of attested FCF, at the current milestone percentage (20/30/40). The percentage steps are constants keyed to cumulative attested FCF, not admin-settable parameters.
• The buyback leg: an independent execution agent calls buybackAndBurn(), which enforces the pre-committed formula on-chain (TWAP-referenced pricing, daily volume caps, and reverts inside blackout windows around disclosures). Purchased tokens are burned in the same transaction. Every burn emits an event; every event is traceable to a published wallet.
• The tender leg: quarterly, openTenderWindow() opens a Dutch-auction tender funded by the Engine's tender allocation, floored at the discount-bounded attested NAV. Holders tender; the auction clears; the contract settles.

Nothing in that flow requires anyone at Napkin to make a discretionary decision, and nothing in it can execute against a stale NAV (Section 7.3).

// SPDX-License-Identifier: BUSL-1.1

pragma solidity ^0.8.26;

/// @title INPKN: ERC-3643-compliant permissioned token, full ERC-20 interface preserved

interface INPKN /* is IERC20, IERC20Permit */ {

    /// @notice Compliance-checked transfer; reverts with a typed reason code on ineligibility.

    function transfer(address to, uint256 amount) external returns (bool);

    /// @notice Identity registry consulted on every transfer (ONCHAINID claims).

    function identityRegistry() external view returns (address);

    /// @notice Swappable compliance module; replaced behind the 48h timelock, token never upgraded.

    function compliance() external view returns (address);

    /// @notice Burns tokens held by the Engine's buyback wallet. ENGINE_ROLE only.

    function burn(uint256 amount) external;

    /// @notice ERC-3643 lost-wallet recovery; issuer-executed, fully event-logged.

    function recoveryAddress(address lostWallet, address newWallet) external;

    /// @notice Emergency pause state (security council, no timelock).

    function paused() external view returns (bool);

}

/// @title INPKNEngine: the non-discretionary Token Cash Engine

interface INPKNEngine {

    /// @notice Records an auditor-attested NAV. Callable only by the oracle adapter.

    function attestNAV(uint256 navPerTokenE18, uint64 asOf, bytes32 attestationHash) external;

    /// @notice Latest attested NAV and timestamp; all consumers MUST enforce staleness bounds.

    function latestNAV() external view returns (uint256 navPerTokenE18, uint64 asOf);

    /// @notice TWAP-referenced buyback within daily caps; burns atomically; reverts in blackouts.

    function buybackAndBurn(uint256 maxSpendStable) external returns (uint256 burned);

    /// @notice Opens a quarterly Dutch-auction tender floored at discount-bounded attested NAV.

    function openTenderWindow(uint256 fundingStable, uint256 floorPriceE18, uint64 closesAt)

        external returns (uint256 windowId);

    /// @notice Engine share of attested FCF in bps; steps at cumulative-FCF milestones. Not settable.

    function engineShareBps() external view returns (uint16);

    event NAVAttested(uint256 navPerTokenE18, uint64 asOf, bytes32 attestationHash);

    event BuybackBurned(uint256 stableSpent, uint256 npknBurned, bytes32 twapRef);

    event TenderOpened(uint256 indexed windowId, uint256 funding, uint256 floorPriceE18);

}

Two implementation details a careful reader will look for. The sNPKN staking wrapper is an immutable ERC-4626 vault, and yes, it overrides _decimalsOffset() (OpenZeppelin's default of zero provides no first-depositor protection) with a +6 virtual-share offset, zero-share reverts, and a burned seed deposit at deployment. And the contracts ship under the Business Source License 1.1, the Uniswap v3 pattern: commercial forking is barred for the license term, then the code goes permissive. We are under no illusion about what BSL does: when Uniswap v3's license expired in April 2023, PancakeSwap forked it within days [35]. Licenses do not prevent forks; they schedule them. Ours buys a multi-year head start, and Section 7.4 explains why the fork threat is mostly beside the point for this asset anyway.

7.3 Oracles and Attested NAV

Let us start with the honest framing: the NAV of a portfolio of private operating companies is an off-chain valuation problem. No oracle creates trust; an oracle transports it. Our job is to make the transport tamper-evident and the source accountable.

The pipeline runs on Chainlink's SmartData suite. The NAVLink pattern already used in production by Fidelity International and Sygnum [51]. The data source is not Napkin's marketing department. It is the attestation output of the transparency stack in Section 8: the audited IFRS consolidation, the quarterly ISAE 3000 examination-level per-company attestations, and the real-time API feeds from portfolio-company banking and accounting systems. The fund administrator and auditor sign the attested figure; the Chainlink decentralized oracle network publishes it on-chain on a heartbeat-plus-deviation schedule; the attestation hash written with each update commits to the underlying documents, which are published on the Proof of Balance Sheets dashboard. Anyone can check that the number on-chain is the number the auditor signed.

Alongside NAV, we wire the Proof-of-Reserve pattern into the mint path with Secure Mint semantics: new issuance requires attested backing greater than or equal to supply plus the mint amount, enforced programmatically [51]. This is the on-chain expression of the NAV discipline rule from Section 5 (no issuance below attested NAV per token) as an invariant rather than a policy.

Staleness is a first-class failure mode

Every consumer of latestNAV() enforces a staleness bound. If the feed goes stale or deviates outside configured tolerance, a circuit breaker freezes NAV-dependent actions (Engine buybacks, tender pricing, primary issuance) while leaving transfers untouched. Holders can always move their tokens; the machine simply refuses to price anything against a number it cannot trust. Combined with the 120-day audit covenant (where a missed audit steps the Engine percentage up against the issuer, not the holders), staleness becomes expensive for exactly one party: us.

Why so much machinery instead of assuming the backing is fine? Because hard-coded 1:1 assumptions are how asset-backed tokens die: every collapsed backed structure (from algorithmic pegs to self-reported real-estate marks) assumed a token-backing relationship no independent party was continuously verifying (Section 8.1 has the autopsy). We ban the assumption at the architecture level: backing is an input (attested, timestamped, checked) or the dependent function reverts.

7.4 The Walled Garden, Used Correctly

Every founder building on open rails eventually asks the same question: what stops someone from forking us? It deserves a real answer, so here is the autopsy of the most famous fork in DeFi history.

In September 2020, SushiSwap forked Uniswap v2's open-source code, bolted on a token, and paid it to anyone who staked Uniswap LP positions. On September 9, the migration pulled roughly $810M (about 55% of Uniswap's liquidity) in a single day; over $1B was staked within ten [28]. The vampire attack won, decisively, for about two weeks. Then it lost the decade. By 2025–26, SushiSwap's TVL sat near $102M, down 98.7% from its ~$8B peak, while Uniswap held $4B+ in TVL and collected roughly $985M in protocol fees from January to October 2025 alone [28]. A 2025 Journal of Operations Management study of vampire attacks found they routinely strengthen incumbents by forcing faster shipping [53]. Forks copy code. They do not copy liquidity network effects, audit history, integrations, brand, or teams.

For $NPKN, the fork threat model is weaker still, because our moat is not code at all. A forker of the NPKN contracts would own a token pointing at nothing: no share of NewCo, no title to the machine or its free cash flow, no audited consolidation to attest, no acquisitions.

Nobody can fork a portfolio of real operating companies, a EUR 500M Vienna-listed bond program, and an audited consolidation.

The precedent is Figure on Provenance: anyone could always fork Provenance's code, and Figure's $17–20B tokenized HELOC book never moved, because the moat is origination volume and legal ownership [54]. Ours is the same shape. This is why we do not hide $NPKN behind a private chain. A walled garden would strangle the token's distribution while protecting nothing that needed protecting.

Where the walled garden genuinely belongs

There is, however, a place in the Napkin architecture where a permissioned chain is not paranoia but correctness: the future RSX settlement layer. Revenue-share settlement between portfolio companies moves exactly the data we should not publish: per-company revenue, waterfall splits, counterparty terms, acquisition-pipeline signals. Commercially sensitive, audience-restricted by nature.

The design is a post-ACP-77 permissioned Avalanche L1 (Evergreen-style PoA). Since the December 2024 Etna upgrade, an Avalanche L1 defines its own validator admission logic through an on-chain ValidatorManager contract, and validators pay the P-Chain roughly 1.33 AVAX per validator per month [55]. The RSX chain runs Napkin-appointed validators: Napkin entities, our auditor, and a banking partner, so the parties with legal duties to see the books are the parties operating the ledger. All validators are Canadian-hosted for data residency, mirroring the Intain precedent, whose IntainMARKETS subnet, servicing $5.5B+ in asset-backed securities, requires verified US-entity validators on US infrastructure [56]. Transaction-submitter and contract-deployer allowlists plus KYB gating control who touches the chain at all.

And because access-control privacy is not cryptographic privacy (every admitted node sees the full ledger) we layer eERC encrypted balances on top: the AvaCloud Encrypted ERC-20 standard (zk-SNARKs plus partially homomorphic encryption) [55], so even in-network participants see only their own flows, while rotatable auditor viewing keys preserve exactly the selective disclosure a Vienna-listed issuer owes its trustees and regulators. An ICM (Interchain Messaging) gateway bridges settled, netted value to public chains when we choose. A walled garden with a controlled gate, not a bunker.

None of this is exotic. Progmat (Japan's largest security-token platform, backed by MUFG) migrated $2.8B of tokenized securities from R3 Corda to a dedicated Avalanche L1 in 2026 [56]. JPMorgan's Kinexys runs tokenized funds across a permissioned Avalanche Evergreen subnet under MAS Project Guardian [56]. Citi ran its Project Guardian FX pricing PoC on private Evergreen subnets [56]. When regulated institutions choose a permissioned settlement venue in 2026, this is the one they keep choosing. We simply arrive with a use case it actually fits.

7.5 Security and Launch Path

Security spend is where token projects reveal whether they expect to exist in five years. Our budget assumes we do.

Dual independent audits. Two full private audits from separate top-tier firms (OpenZeppelin and a Spearbit/Cantina team) with the second run on frozen post-remediation code. For a scope spanning a permissioned token, compliance layer, Engine, async issuance flows, and CCIP integration (bridge-adjacent complexity) we budget $150–300K across both engagements, plus a public audit competition on the frozen commit with a $30–80K prize pool. Booking queues at these firms run 4–12 weeks, which is why auditor slots are reserved at spec freeze, not after the build. Alongside human review, a Certora/Halmos formal-verification suite proves the invariants that matter: cross-chain supply conservation, the Secure Mint bound, no transfer path bypassing compliance, ERC-4626 share-price monotonicity, pause completeness, and no upgrade path outside the timelock.

Bug bounty, live before the token. An Immunefi program at the institutional norm (10% of funds at risk for critical findings, capped at $1M at launch and scaling toward $5–10M with TVL) published and funded before mainnet deployment, not after the first incident. Immunefi has paid out $112M+ across its programs [57]; we would rather pay a researcher than fund an attacker's exit.

Engineering discipline. Foundry CI with >90% branch coverage, fuzz and invariant testing (Echidna/Medusa) on supply and compliance choke points, and Slither static analysis on every commit. Deployment is deterministic CREATE2 from a clean deployer, bytecode-verified on every chain, with all roles transferred to the Safes and timelock and the deployer renounced before TGE. Monitoring runs on Hypernative and Tenderly with a rehearsed auto-pause playbook. And one line most papers would not print: Napkin's operations run on 62 AI employees, so every AI-agent-initiated on-chain action sits behind the Fireblocks policy engine and Safe thresholds. No autonomous agent ever holds a complete key. That constraint is written into the audit scope and the threat model as a first-class item.

The launch path is a staged ratchet, with each stage gated on the previous one: dates slip before gates do.

End to end, that is roughly 7–9 months from spec freeze to unrestricted mainnet, and $250–450K all-in for security including bounty funding and monitoring. We consider that number a feature. A token claiming a multi-billion-dollar destination on a five-figure security budget is telling you its own expected value: ours is priced for the destination we actually intend to reach.

The bug bounty goes live before the token does. The caps come off after the audits: never before.

Proof of Balance Sheets: Transparency & Oracles

A token whose value derives from private-company financials has exactly one existential dependency: can anyone outside the issuer verify the numbers? If the answer is no, every mechanism in Section 5 is "trust me" with extra steps. This section is our answer, and it ships before the token does.

8.1 Why Self-Reported Marks Are Disqualifying

First, the autopsy that shapes everything here. RealT sold fractional tokens on Detroit rental homes to roughly 14,000 mostly European retail investors. In 2025-2026 it emerged that RealT had sold $2.72M of tokens on 39 homes it never owned (the deals had simply not closed) while paying "dividends" on properties with no tenants and accumulating 400+ uninhabitable properties, culminating in the largest nuisance-abatement lawsuit in Detroit's history and a portfolio liquidation (court filings, 2025-2026) [58].

Note what did not fail: the blockchain. The tokens transferred flawlessly. The smart contracts executed perfectly. The operating company simply lied about what it owned, and no independent party was positioned to catch it. That is the entire lesson of the RWA cycle's worst failure: self-reported marks are not a weak form of verification. They are no form of verification. For a portfolio scaling from a young deal count toward 10,000, one RealT-style discovery would end $NPKN permanently. So we built the verification stack as if that discovery attempt happens every quarter. Because it will: short sellers already target exactly this surface.

8.2 The Three-Layer Verification Stack

Layer 1. One audited IFRS consolidation. Every acquired company sits inside a single IFRS consolidation perimeter under Napkin Group AG, audited annually by a named top-10 firm under the 120-day covenant of Section 6.4. One perimeter, one group audit with standard component-materiality scoping. Not 10,000 separate audits, which would be theater, but the structure global audit practice actually uses to make a thousand subsidiaries verifiable. Ownership of each acquired entity is evidenced by registry extracts, hashed on-chain. The RealT failure mode (tokens on assets never owned) is structurally closed at this layer.

Layer 2: Quarterly ISAE 3000 attestation of the deals. Every quarter, an independent firm performs an examination-level attestation (ISAE 3000) [36] of exactly the numbers that drive the Engine: per-company revenue, EBITDA, and cash, consolidated FCF, and Qualifying-Acquisition status. Examination level (the standard that supports an opinion) not a compilation, not agreed-upon procedures. The attestation lag is disclosed (target: 45 days or less), and all Engine execution and tender pricing runs exclusively against last-attested values.

Layer 3: Real-time feeds and the on-chain NAV oracle. Portfolio companies pipe operating data (Stripe, banking, and accounting APIs) into the Proof of Balance Sheets dashboard in near-real-time, and a Chainlink oracle (following the SmartData / Proof-of-Reserve pattern already live for WisdomTree and others) [51] publishes attested NAV per token on-chain, with the signed attestation report hashed alongside it. The Engine's smart contracts execute only against auditor-signed data.

One honest boundary, stated before a critic states it for us: the oracle transmits; it does not verify. No oracle can audit a private company's EBITDA. Verification lives in Layers 1 and 2: human auditors, examination standards, legal liability. Layer 3 makes the verified numbers unfakeable in transit and permanently public. Anyone who tells you their oracle "verifies" real-world financials is describing Layer 3 while hoping you don't ask about Layers 1 and 2. We also disclose that our internal accounting stack is substantially AI-operated (the 62-agent workforce of Section 4) which is exactly why the internal-controls audit scope explicitly covers the AI-agent pipeline. First-of-kind disclosure, and we would rather own it than have it discovered.

8.3 The Dashboard Ships Before the Token

The Proof of Balance Sheets dashboard (live NAV per token, FCF per token, per-deal attested metrics, Engine wallet activity, the burn tracker against the 500M terminal supply, every attestation report) goes live before TGE. Not on a roadmap. Before. The durable launches of this cycle shipped product before token; the hype-first ones (Plasma's -90% chart is canonical [30]) shipped promises. You will be able to watch the portfolio breathe for months before you can buy a claim on it. If the numbers underwhelm you, don't buy. That option (informed refusal) is the whole point of transparency.

8.4 Market-Maker Transparency

Liquidity provisioning is where clean tokenomics goes to get quietly dirty, so we adopt the strictest 2026 disclosure norms voluntarily:

• Our market makers are named in this document's launch supplement, not hidden behind NDAs.
• Retainer-model contracts only: flat fees against spread, depth, and uptime KPIs. No loan-plus-call-option structures, whose mispriced strikes create hidden sell pressure and phantom float. If any token loan ever exists, its size, term, and strike appear in the supply tables, and loaned tokens are counted as circulating: always.
• A contractual ban on wash trading, volume guarantees, and price targets in every MM agreement. The DOJ's Operation Token Mirrors prosecutions (CLS Global's conviction, Gotbit's plea) made wash-trading-as-a-service a crime [59]; we make contracting for it a breach.
• A monthly liquidity report (spreads, depth, venues) published alongside the buyback attestation.

Quant desks now reconstruct MM inventories from on-chain flows in an afternoon. Being caught not disclosing is strictly worse than disclosing. So we disclose.

8.5 The MAR Overlay: Already Regulated, Gladly

Because the EUR 500M bond program is listed in Vienna, Napkin is already subject to the EU Market Abuse Regulation [60] (inside information rules, insider lists, disclosure discipline) today, before any token exists. Most crypto issuers would call that a liability. We call it the asset it is: the buyback program of Section 5.5 is built to MAR's equity safe-harbor logic precisely because we live under MAR; a named human compliance officer (not an AI agent) owns insider-list and blackout-window discipline; and a serial acquirer's continuous stream of inside information is handled by the same procedures European listed companies have run for decades. We were a regulated capital-markets issuer before we were a token issuer. The token inherits that spine. It does not have to grow one under enforcement pressure.

Proof of Balance Sheets is not a marketing layer. It is the product: attested numbers, or no Engine.

Corporate & Regulatory Architecture

Most token projects treat securities law as the enemy. We treat it as the moat. $NPKN is a security by definition (a tokenized common share of NewCo) because common equity is the only structure under which the sentence "the holders own the machine" is legally true rather than marketing fiction. Every credible retail path on Earth in 2026 runs through securities law, not around it. So we run through it, in five jurisdictions at once.

That multi-entity design is also our decentralization argument, stated in language a regulator can love: no single regime, bank, or intermediary is a single point of failure for the network. This is not about escaping governments. It is multi-jurisdiction redundancy. The same logic that makes a distributed ledger worth running in the first place, applied to the corporate stack that stands behind it.

9.1 The Issuer Stack

First, the current state, plainly. As of the time of writing, the group parent is Napkin Inc. (Canada) (the umbrella holding company and IP owner) with Napkin Global Ltd. (ADGM) as an affiliate, Napkin USA Inc. (Florida) as the U.S. operations holdco, NPKNSPV1 Inc. (Florida) as the SPV holdco for the construction roll-up, Skynet (India) as the engineering arm, and Napkin LATAM/Levector serving Latin America [3]. The stack below is the target token architecture that the Conversion of Section 6.7 builds: NewCo, a British Columbia corporation, becomes the group parent and the issuer of the tokenized common shares, acquiring Napkin Inc. and its subsidiaries through the Arrangement, subject to bondholder change-of-control consent. One honest note: the ADGM affiliate remains in the group, but the token program is deliberately structured outside the Middle East. And one deliberate alignment: the group parent is already Canadian, so the Arrangement continues a Canadian parentage rather than inventing one, in the founder's home province.

NewCo: British Columbia, Canada

The apex of the stack arrived in Version 5.0, and its domicile is doing real work. NewCo is a British Columbia corporation because the Arrangement is BC law and because a share-for-share exchange into a Canadian corporation preserves the tax rollovers embedded in the legacy table: Section 6.5 has the full logic, Section 6.7 the mechanics. The Arrangement itself relies on the reorganization exemption: securities issued to existing holders under a court-approved statutory procedure require no new prospectus, with the fairness opinion and the court order protecting dissenters (NI 45-106 s.2.11) [64]. New money never uses that exemption; every raise runs through the compliant lanes below. And the domicile compounds with the rest of this chapter: the arrangement court, the founder's residency and tax planning, and the uncapped BC offering-memorandum lane all sit in the same province as the issuer.

Napkin Group AG: Zug, Switzerland

The portfolio holding company sits in Zug because Switzerland offers the deepest legal certainty for tokenized securities anywhere: the DLT Act has enabled ledger-based securities with insolvency protection since 2021, and the banking relationships crypto companies elsewhere still beg for (Sygnum, AMINA, both full FINMA banking licenses) are native infrastructure here [61]. In the target stack, Napkin Group AG holds the acquired-company portfolio, the AI-operations IP, and the equity of the operating entities below it, under NewCo. Keeping the portfolio holdco in Zug (rather than concentrating everything in any single regime, the ADGM affiliate included) is concentration-risk management: no single regime dependency.

Napkin Securities AG: Vaduz, Liechtenstein

This entity made lawful EEA retail real in every prior version, and Version 6.0 retains it for exactly that job: EU distribution. Because a tokenized share is a MiFID II financial instrument, it sits outside MiCA entirely (Art. 2(4)(a)) and is offered under an EU Prospectus Regulation prospectus approved by FMA Liechtenstein, which passports to all 30 EEA states [46]. One approved prospectus, thirty countries, genuine retail access for a security token: that math survives the equity recast unchanged. What counsel still decides is the wrapper (Section 6.5): a direct passported prospectus for the tokenized BC share, or an EEA depositary instrument issued by this entity that wraps the share one-to-one, with Liechtenstein's TVTG (the world's first comprehensive token-container law) making that token the legal carrier of the wrapped right [45]. Intent plus options, decided in the prospectus process and published in the token documents. A PRIIPs Key Information Document accompanies the retail offer either way [46], and the group's Vienna bond gives Napkin existing EEA disclosure history that materially shortens FMA approval.

Napkin Markets Inc.: Delaware

The US tranche runs as Reg D 506(c) [62] (general solicitation to verified accredited investors) with Securitize, the first FINRA-approved end-to-end on-chain broker-dealer, acting as SEC-registered transfer agent, distribution channel, and ATS [63], the same agent architecture that already keeps the register for SEC-qualified tokenized common shares [107]. Securitize also manages the Exchange Act 12(g) trap: 2,000 record holders (or 500 non-accredited) plus $10M in assets forces full reporting-company status [62], so holder-count management via the transfer agent is mandatory, not optional. The retail path is phased and honest: Reg A+ Tier 2 (up to $75M per 12 months to non-accredited investors) [62] as the compliant beachhead, and the SEC's Innovation Exemption under Project Crypto (the 12-to-36-month sandbox for issuing and trading tokenized securities on-chain, released in mid-2026 [62]) as the scaling lane. We are application-ready. We do not premise anything on the CLARITY Act passing.

Napkin Operations (Canada) Ltd.: Victoria, British Columbia

This is a real operating company (jobs, engineering, SR&ED credits) in the founder's home province. It also unlocks a lane almost nobody in crypto knows exists: British Columbia retains the original, uncapped offering-memorandum exemption under NI 45-106 s.2.9(1). Any purchaser, any amount, no eligible-investor requirement: deliver a compliant offering memorandum (Form 45-106F2) and a signed risk acknowledgement, and a broader-than-accredited raise is lawful [64]. Ontario, Alberta, and Quebec cap the same exemption at $10,000 for non-eligible investors and $30,000 for eligible ones [64]. BC does not. Add Canada's 4-month resale hold (NI 45-102) [64] versus the 12-month US Rule 144 lock [62], and BC is a genuine Canadian retail lane, not a footnote. We engage the BCSC's Tech Team through pre-filing meetings rather than launch-then-negotiate. Founder residency and cross-border tax planning (CRA departure tax under ITA 128.1(4), s.94 deemed-resident-trust rules, FAPI attribution) [65] is flagged openly and handled with Canadian tax counsel: see Section 11 for the RSX reserve structure.

Napkin Foundation: Cayman Islands

A memberless, orphaned Cayman Foundation Company will steward the future RSX protocol and its ecosystem treasury. The market-standard wrapper exchanges and counterparties expect, strengthened by the 2025 VASP Amendment that narrowed "issuance of virtual assets" to public sales for consideration [66]. One separation is absolute: the Foundation never touches the security token. $NPKN lives entirely inside the securities perimeter; the Foundation exists for the protocol era.

9.2 The Vienna Bond: an Honest Footnote

The EUR 500M ETN program (full terms in Section 4) stays listed in Vienna, and the EU Market Abuse Regulation applies to us because of it. But we will not oversell it: the Vienna MTF is the exchange-regulated segment, listing by information memorandum, no prospectus, no ongoing financial-statement obligation [24]. That is precisely why the $NPKN token terms carry contractual disclosure covenants that exceed the venue's requirements. The full stack, penalty inversion included, is in Section 6.4. We do not borrow credibility from a listing segment. We write it into the contract.

One practical note on the rails, restated from Section 4 because it matters to every jurisdiction in this chapter: the program raises euros and the machine deploys dollars, where the bulk of the acquisitions close. International deals settle in whatever currency the seller values, and stablecoin settlement is welcome the day a seller prefers it. The rails work for us, not us for the rails.

9.3 Secondary Markets

• EEA: 21X in Frankfurt. The first licensed DLT Trading and Settlement System, live since September 2025, with direct retail access and no broker intermediation [49]. The European Commission's April 2026 proposal to raise the DLT Pilot platform threshold from EUR 6B to EUR 100B removes the scale ceiling [49].
• US: Securitize's ATS for accredited secondary trading, expanding as the Reg A+ and Innovation Exemption tranches open.
• Canada: a registered marketplace/CTP path later. We are honest about this: no Canadian venue lists exempt-market security tokens at scale today. BC purchasers get the 4-month hold and our published liquidity roadmap, not a fantasy.

One clarification, stated once so it never needs restating: the group still intends an exchange listing in time, a regional exchange first, then NASDAQ after a proven capital-allocation track record. Since Version 5.0 that is no longer a separate track: it would be a listing of the same class of common shares the token represents, adding venues rather than creating a second instrument. Nothing about $NPKN is contingent on any exchange listing occurring.

9.4 We Read the Enforcement Actions So You Don't Have To

In March 2025, BaFin prohibited the public offer of Ethena's USDe in Germany, froze reserves, levied a EUR 600,000 coercive fine, and forced a supervised 42-day redemption wind-down and EU exit, and separately flagged the staked, yield-bearing wrapper as a suspected unregistered security (BaFin; DLA Piper, 2025) [47]. That product was less security-like than ours. The lesson: a yield-bearing instrument offered to EU retail without authorization gets shut down, every time, and the staking wrapper goes down with it. That is why our staking yield lives inside the prospectus perimeter from day one, paid in stablecoins from attested cash flow rather than emitted tokens, and why we pre-filed instead of launching and negotiating. Prospectus-first is not a compliance tax. It is the strategy: the difference between a token lawfully marketed to 30 countries of retail investors and one that gets a redemption plan imposed on it.

9.5 Ethical-Finance Alignment

One property of the capital architecture deserves stating on its own, because it was designed in from the start, not retrofitted for a market: the structure is built to align with ethical-finance principles by construction, and it is designed to satisfy recognized ethical-finance certification standards. The boundary is stated plainly: formal certification requires a qualified supervisory board, and we do not claim it. The intent is inclusion: rails two billion more people can stand on without violating their convictions.

• Riba-free by construction: capital partners earn a contractual share of actual distributable cash flow, in proportion to the capital they provided; zero cash flow means zero return; no instrument in the target architecture pays fixed interest.
• Asset-backed: the secured notes are certificates sold to investors, not loans borrowed.
• Gharar minimized: earn-out conditions are extinguished by full settlement at the Release.
• Sector-screened: the acquisition screens exclude prohibited sectors from the start.

Go-to-Market: Product Before Token

Every durable winner of 2024-2026 (Hyperliquid, Ethena, Aster) had a working product with real usage before the token existed. The worst-performing categories were the ones with nothing underneath the narrative. Napkin has 15 completed acquisitions, a consolidated cash-flowing operating portfolio, and a shareholder-reporting track record. Launching on story alone would waste the one advantage almost no token has ever had. So the sequence is non-negotiable: product ships first, proof ships second, token ships third.

10.1 The Five Phases

Phase 0 is the tell. Before a single token exists, the Proof of Balance Sheets dashboard goes live (Section 8.3): every acquired company as a deal (revenue, EBITDA, cash, audit status) updating in public. It is our proof-of-reserves, except the reserves are operating companies. If we cannot ship the dashboard, we have not earned the right to ship the token.

10.2 Deal Scout Season 0

Points programs died of their own cynicism. Blast farmed deposits for 17 months and collapsed at TGE [67]. EigenLayer hid its allocation math and triggered a revolt [67]. Ethena, the counterexample, ran a six-week season that rewarded actual product usage, and its product kept growing after the token launched [67]. We copy the winner.

Season 0 is one capped season, 8-10 weeks, and then it ends. Points are earned only for real economic contribution to the M&A machine:

• Submitting SMB acquisition targets that pass diligence. The scarcest input in small-cap M&A.
• Referring sellers of small companies directly.
• KYC-verified pre-registrations for the compliant sale tranches.

Closed deals pay real bounties: 0.5% to 1.5% of closed enterprise value, in vested NPKN. Standard VC-scout economics, trivially cheap against the 2-8% bankers charge on SMB deals. The exact points-to-NPKN conversion and the total Season 0 allocation are published before the season opens, not after: opacity is what caused the EigenLayer revolt, and we will not repeat it. The leaderboard is reputation-weighted and runs on our own infrastructure. The X API shutdown of January 2026 proved what single-platform dependence costs [32].

The capital mechanics around those deals are published, not improvised. Pledges appear on the dashboard with their pledged values before any capital moves: the commitment is visible first, the cash second. Every pledge receives a definitive answer within 40 days, and that deadline is a covenant in the documents, not a service target. And batch closes run through an accretion gate: a batch of deals executes only when the attested numbers show it accretive per share, so the machine grows value, not just size.

And here is the moment nobody else can copy: we intend to close one to two community-sourced acquisitions during the season and publicize the paid bounties while it is still running. Other projects hand out points for retweets. We hand a community member a scout fee for sourcing a real company that our machine actually bought. That is the marketing. The flywheel is the content.

10.3 The Community Round

Phase 2 runs a strategic round through an Echo/Legion-style community platform: the channel that legitimized the ICO again, with MegaETH drawing $1.39B in bids and Yield Basis 98x oversubscription [68]. Allocation is merit-scored, not first-come-first-served. The cap is deliberately modest: an oversubscribed small raise is worth more than a maximal one. And the discipline that matters most: strategic participants get identical vesting to every other purchaser, per the rule of Section 5.2. No side letters, no better terms, disclosed in full. The market learned to reconstruct insider terms on-chain; we simply publish them.

Pricing follows the one lesson venue legitimacy cannot teach: Monad sold through the most legitimate platform in crypto at a $2.5B FDV and broke 33% below sale price within a month [68]. We price against conservative attested portfolio NAV plus pipeline (a number we can defend with an auditor's signature) never against a headline FDV print.

10.4 TGE Structure

TGE is gated, not dated. Four gates, each independently verifiable, all green before the first token trades:

• G1: The Conversion holds. Phase 1 Trim complete, with the Nobody Loses Covenant honored and the results disclosed; Phase 2 Arrangement effective, court order granted and bondholder change-of-control consent in hand (Section 6.7).
• G2: The proof is live. Proof of Balance Sheets running on attested data feeds, per Section 8.
• G3: The lanes are open. Prospectus and exemptions in force; transfer agent and trading venues live.
• G4: The four numbers are published. Raise size, pricing basis, TGE NAV per share, and the milestone thresholds, in the token documents.

We don't TGE on promises. We TGE on a court order, an attestation, an approval, and four published numbers.

One clarification so the two rulers never blur: the gates govern whether the token launches; the cumulative attested-FCF milestones of Section 5.4 still govern when the Engine steps up. Both are hard-currency tests, and neither is negotiable.

The structure behind the gates:

• At least 40% genuine float at TGE. The community and public distribution tranche is the float. No accounting games (Section 5.3 has the dataset).
• Distribution through the three compliant lanes simultaneously: EEA retail via the passported prospectus on 21X, US accredited via Securitize, Canadian retail via the BC offering-memorandum exemption.
• Named market makers, retainer-model contracts only, loaned tokens counted as circulating, contractual wash-trading ban in every agreement. The full policy is Section 8.4.
• The Engine switches on at TGE, funded by existing portfolio cash flow. Even modest day-one buybacks establish the mechanical link between the business and the token.
• Full unlock calendar on-chain from day one, with the lock rules of Section 5.2 enforced in the contracts.

10.5 Market Integrity and the Liquidity Reserve

A launch is won or lost in its order book. We carve a Liquidity and Listings Operations reserve of up to 5% of supply, unlocked at TGE, as an authorized ceiling rather than a deployment plan, and we publish its full decomposition, which almost no issuer does. Every liquidity covenant is denominated in percent of free float, not percent of supply, because free float is what actually trades.

• Market-maker inventory: roughly 1.2% of supply held by our market makers as working inventory, loaned under retainer terms and counted as circulating, never as collateral.
• Cash working capital: 500,000 to 1.3 million dollars of NewCo balance-sheet cash in year one, at the practitioner norm of roughly 60% stablecoin and 40% token, disclosed as its own line.
• Protocol-owned liquidity: 1 to 2% of free float seeded into on-chain pools where lawful, managed by the retainer market maker.
• Stabilization and undeployed reserve: the greenshoe over-allotment described below, plus a disclosed remainder held back for future venues.

The hard covenant is simple: no single market maker ever holds more than 2% of free float, and all market makers combined never more than 4%. With a 40%-plus float that is stricter and more honest than a raw percentage-of-supply figure, and it makes it structurally impossible for one counterparty to dominate the book. The NAV discipline rule of Section 5 governs primary issuance of new shares, not the market makers' second-by-second quoting.

10.6 Market Makers: Retainer Only, Named, and Disclosed

We contract market makers on retainer only: flat fees against spread, depth, and uptime targets. We do not use loan-plus-call-option structures, whose mispriced strikes create hidden sell pressure and phantom float. Our policy is written to exceed the strictest exchange standards, including the 2026 market-maker guidelines, even where we are not bound by them.

• The standard: named market makers, verified operating entities, no options or warrants on the token, no profit-share, no discounted purchases, no side investment, no non-disclosure agreements, real-time dashboards, and termination without penalty on short notice.
• Targets: spread within 50 basis points tightening over the first 72 hours, at least 100,000 dollars of depth per side within 1% of mid, and at least 98% two-sided uptime, with partitioned venue mandates.
• Oversight: market makers win a competitive request for quote, not a bilateral deal, and an independent third-party monitor verifies their behavior and feeds the monthly report.

10.7 The Stabilization Program: A Real Greenshoe

Because a tokenized share on a licensed venue is a financial instrument under EU market rules, it can do something no ordinary token launch can: run a disclosed, legal price-stabilization program, the same greenshoe every regulated IPO uses. We intend to be the first token launch with an actual greenshoe, operated within the market-abuse stabilization safe harbour rather than through any derivative in the dark.

• The terms: a stabilization window of at most 30 days, purchases never above the offer price, an over-allotment option of up to 15% of the offer, a naked short capped at 5%, a named licensed agent, and every stabilization trade disclosed within seven trading sessions.

This is the compliant, in-daylight answer to the instinct that a launch needs downside support. Founders do not hedge, the treasury does not short its own equity, and price support happens through a disclosed program under rules written for exactly this purpose.

10.8 Launch Sequencing and the Token Movement Calendar

The order of operations is a published rule, not a judgment call made under pressure. Distribution closes first. Then the stabilization agent may operate for up to 30 days. Only then does the buyback Engine activate, as a separately disclosed, pre-committed program. The Engine never runs during the offering or the stabilization window.

• Venue sequencing: open the deepest regulated venue first to print a reference price, the security-token venue second, and any crypto-native venue last, never a simultaneous big-bang.
• Token movement calendar: every treasury or market-maker transfer of tokens to a venue is announced before it happens. The silent pre-launch wallet transfer is a named cause of failed launches; we remove it.
• Transferability calendar: each resale restriction, by tranche and jurisdiction, is published with its exact expiry date at TGE.
• Unlock discipline: 30 days before any cliff we publish the schedule, pre-place large unlocking holders through compliant over-the-counter blocks, and report a buyback-coverage ratio against a proven benchmark.

10.9 Treasury Policy

A token dies when its treasury is a single volatile asset. Ours is built to survive a bear market and to never depend on selling the token to make payroll.

• Runway: 18 to 24 months of consolidated operating expenses held in fiat and stablecoins at the holding company, independent of the token price.
• Native holdings: treasury NPKN is strategic buyback inventory, and any diversification runs on pre-committed, mechanical, publicly disclosed rules.
• The bright line: payroll and operations are never funded by token sales, and the issuer never hedges, shorts, or writes derivatives against its own equity token.

10.10 Ongoing Disclosure and the Drawdown Protocol

The transparency we promise before TGE continues after it, on a fixed cadence, in good markets and bad.

• Monthly liquidity report: per-venue spread against target, depth at 0.1% and 1% from mid, two-sided uptime, per-market-maker token and stablecoin balances with on-chain wallet attestation, and loan utilization.
• Quarterly holder reports: equity-grade investor relations, a category almost no token provides, including the buyback-coverage ratio and attested per-share metrics.
• Drawdown protocol, in writing: if the price falls 30% from reference, a letter from the chief executive follows within 48 hours, and the monthly report never skips a cycle in a drawdown.

10.11 What We Will Never Do

Six commitments, in writing, before the first token exists. Hold us to every one of them.

• No low-float / high-FDV launch. The 84.7% / -71% failure pattern of Section 5 speaks for itself, and the biggest FDV prints did worst [26].
• No multi-season points farming. Blast's 17-month farm ended with TVL down 97% and daily users down from 180,000 to 3,800 [67]: mercenary capital leaves the moment points stop.
• No paying for Tier-1 listings. PUMP fell despite simultaneous Binance and Upbit listings [43]; Hyperliquid reached a top-11 market cap with neither [31]. Listings are liquidity milestones, not price strategy.
• No pre-selling RSX economics. Prospective, task-conditioned distributions trigger securities treatment under SEC/CFTC Release 33-11412 (March 2026) [21].
• No celebrity tokens. TRUMP and MELANIA cost retail roughly $4.3B; LIBRA erased ~$4B in hours [32]. A founder's profile is not a substitute for disclosed economics.
• No discretionary buybacks on insider information. We are a serial acquirer under MAR [60]; every buyback runs on a pre-committed formula through an independent agent with blackout windows. The issuer never times its own tape.
• No founder derivatives. Founders never hedge, lend, collateralize, or synthetically short their holdings. Founder liquidity happens once, in daylight, through the disclosed Dial.
• No treasury shorting of our own token. The issuer will not take a derivative position against its own equity. Downside is managed by the greenshoe, the Engine, and the fiat buffer, in the open.

RSX: The Grand Vision

Everything in this section is future tense, and we mean that legally as much as grammatically. Nothing here is offered, priced, or promised. This is what we intend to build after $NPKN has done its job.

Two clocks run through this paper, and this section keeps them honest: RSX is not Era 5's reward, unveiled once the portfolio has done its work; its settlement and verification substrate is built from Era 1 onward, on our own companies, which is what the pilot-network language of Subsection 11.10 already describes. The ambition is rails for rails: the settlement and verification substrate that other venues, stablecoins, and exchanges sit on. We never claim the ocean's flows; we only carry the ocean's carriers.

11.1 The Structural Thesis

Financial systems do not merely reflect inequality; they manufacture it. That is the central argument of Engine of Inequality (Petrou, 2021) [86], and it is the premise RSX is built on. Who gets capital is decided less by merit than by architecture, who may list, who may underwrite, who clears, who takes the spread. Change the architecture and you change the outcome. RSX does not fight markets. It redesigns market access.

This is not a political program. It is protocol design. The subsections that follow are the thesis, implemented. Structural, not political.

"This is not a redistribution model. This is a participation model.": founder's working papers, 2026

11.2 A Third Asset Class: The Revenue Share Agreement

The instrument RSX is designed to list is the Revenue Share Agreement: a standardized contract selling a defined percentage of a company's future revenue, with caps, durations, and triggers. Neither debt nor equity. A third asset class, royalty economics applied to all commerce. One worked example carries the idea. Alice's sock store takes a $50,000 advance against future sales; from that moment, every $20 pair of socks routes $2 to her backers the instant it sells, until they have received $150,000 (a 3x cap) and the stream switches off. No dilution, no covenants, no personal guarantee; if sales slow, remittances slow with them. The obligation breathes with the business. The structure sits deliberately adjacent to profit-participation finance, including Sharia-compliant forms, and one standardized, framework-sourced RSA rulebook governs every listing.

The Behavior Already Exists

RSX does not have to create the behavior of selling revenue. Businesses already sell roughly EUR 4.0 trillion of receivables every year through factoring and receivables finance (FCI World Industry Statistics, 2025) [87] (an annual flow, not a stock) while about $2.5 trillion of trade-finance applications are rejected annually, a measured backlog of unmet demand (Asian Development Bank, 2023) [88]. Packaging cash flows into tradeable paper is likewise the financial system's core industrial process: roughly $2.0 trillion of mortgage- and asset-backed securities were issued in the United States in 2024 alone (SIFMA, 2025) [89], a lineage running back to the $55 million securitization of David Bowie's royalties in 1997 [90].

The digital-native segment shows where this is heading, and what it is missing. Revenue-based financing platforms Wayflyer, Clearco, and Capchase have deployed roughly $10 billion of cumulative capital against future revenue (company disclosures, 2025) [91]. Pipe, branded the Nasdaq for revenue, became the fastest fintech to a $2.0 billion valuation with some $7 billion of annual recurring revenue connected to its platform (TechCrunch, 2021) [92], then its founders departed amid questions over underwriting quality, and the marketplace retreated into embedded lending (Forbes, 2022; Forbes, 2024) [92]. Pipe proved the demand, then proved the constraint: a revenue marketplace lives or dies on the quality of its revenue verification, exactly what the attested-financials stack of Section 8, generalized in Subsection 11.4, exists to answer.

Appetite is proven at every scale. Over $8 billion of music catalog transactions closed in 2024, larger deals averaging 16.1x multiples (Shot Tower Capital via Billboard, 2025) [93]; Blackstone took the Hipgnosis catalog private at roughly $2.3 billion of enterprise value and led a $1.47 billion securitization of it (Billboard, 2024; Blackstone, 2024) [94]. Hipgnosis had first collapsed as a listed fund on opaque, contested valuations; institutional appetite and the case for attested valuation, in one story. And the rails are operational fact: roughly $13 billion of active tokenized private-credit loans were outstanding on-chain by early 2026 (rwa.xyz, Q1 2026) [50], the largest originator an SEC-registered, Nasdaq-listed company (Figure Technology Solutions prospectus, 2026) [54]. None of these figures may be summed (they monetize overlapping receivables) but together they settle the point. Revenue trading is a multi-trillion-a-year behavior running on decades-old plumbing; every experiment proves the demand, and every failure isolates a missing piece of infrastructure. RSX universalizes an existing behavior on neutral rails. It does not invent one.

11.3 Settlement at the Source

The atomic mechanism of RSX is the split at the point of sale. Every revenue event is designed to fan out, at transaction time, into investor, talent, company, and protocol remittances: batched off-chain for cost, settled on-chain for finality. Stakeholders are paid from the revenue event itself, not from invoicing and reconciliation after it: Alice's backers do not wait for her quarterly report, they are paid by her checkout. This diagram has existed in RSX's architecture since 2022; the industry has only now built rails that make it deployable at scale.

Settlement at this granularity moves exactly the data a public chain should never publish: per-company revenue, waterfall splits, counterparty terms. That is why the RSX settlement layer is designed as a permissioned Avalanche L1 with encrypted balances: validators appointed from Napkin entities, our auditor, and a banking partner, Canadian-hosted for data residency, with rotatable auditor viewing keys preserving the selective disclosure a regulated issuer owes its trustees and supervisors. Section 7.4 specifies the full architecture. The validator economics, the eERC encryption model, and the institutional precedents (Progmat, JPMorgan Kinexys, Citi, Intain) that make this a proven pattern rather than an experiment [56]. The walled garden protects commercial data on RSX; it never touches $NPKN, which lives on public rails where a security token belongs.

11.4 The Trust Layer: Continuous Underwriting

Pipe's failure is RSX's design requirement. A market in revenue claims is only as good as its knowledge of the revenue, so underwriting is native infrastructure: live oracle feeds from accounting, banking, and commerce platforms are designed to score every listed stream continuously, pricing on observed cash flow in the present rather than a credit bureau's photograph of the past. This is what makes market-dictated RSA pricing possible, and what makes RSAs tradeable at all. For Napkin the layer is not speculative: it is the Proof of Balance Sheets stack of Section 8, built for the Napkin portfolio and generalized to every future listing. The exchange's trust layer will have run in production, on our own companies, before the first external listing goes live.

11.5 Market Structure: The Exchange the Other 99.98% Never Had

"Everyone has the right to IPO.": founder's working papers, 2025

The Macro Play

Sizing this market honestly requires layers: the numbers live in different units (flows, revenues, market values) and confusing them is how white papers die.

Context first: global payments moved roughly $2.0 quadrillion of value across 3.6 trillion transactions in 2024 (McKinsey Global Payments Report, 2025) [95]. Most of that is interbank, FX, and securities flow RSX does not address. The quadrillion is the ocean the exchange swims in, not our market, and we will never claim it as one.

The addressable layer is the commerce whose receipts could back a revenue-share instrument. B2B payments ran roughly $89 trillion in 2024, projected to reach $124 trillion by 2028 (Juniper Research, 2024) [96], home of the invoice-based, contract-backed cash flows that are the raw material of revenue shares. Consumer commerce adds about $27.6 trillion of annual card volume: Visa carried $17.0 trillion in fiscal 2025 (Visa Form 10-K, 2025) [97], Mastercard $10.6 trillion (Mastercard Form 10-K, 2026) [98]; $6.4 trillion of that is retail e-commerce (eMarketer, 2025) [99], already digital and API-observable, the most instrument-ready slice; roughly 300 million installed POS terminals (Nilson Report, 2024) [100] are where offline revenue becomes observable. These pillars overlap: e-commerce rides card rails, commercial cards overlap B2B, so we quote a range, never a sum. Call it on the order of $100 to $150 trillion of annual revenue flows that could, in principle, carry a claim.

The beachheads are digitally native, recurring revenue where integration is an API call: roughly $300 billion of annual SaaS spending (Gartner, 2024) [101], contractually recurring and metered, the highest-quality collateral class for revenue shares; a creator economy near $250 billion in 2023, projected to approach $480 billion by 2027 (Goldman Sachs Research, 2023) [102]; 154 to 435 million online gig workers, each a platform-verified income stream (World Bank, 2023) [103]; and embedded finance forecast to exceed $320 billion of annual revenue by 2030, roughly $150 billion of it from small businesses (BCG, 2024) [104]. A measure of how willingly commerce platforms adopt finance inside their rails, which is structurally RSX's category.

Now the claim, stated with discipline. Flows are the universe. The $2.5 trillion of annual payments-industry revenue those flows generate (McKinsey Global Payments Report, 2025) [95] is the fee pool actually for sale, and RSX competes for a slice of it, priced in basis points on settled flow. Visa's $40 billion of revenue on $17 trillion of volume (roughly 23 basis points (Visa Form 10-K, 2025) [97]) is the benchmark for what a settlement network earns per dollar it touches; settle even a fraction of one percent of those flows at single-digit basis points and you have a very large business, one designed per Subsection 11.7 to route its take away from entrepreneurs. Feasibility is settled too: adjusted stablecoin volume reached $10.9 trillion in 2025, up 91% year over year (Visa Onchain Analytics, 2025) [105], approaching Visa's own $14.2 trillion of annual payments volume [97]. Blockchain rails already settle at card-network scale.

The Venue

Against all of that stands the incumbent: roughly $190 billion of annual IPO proceeds across about fifty thousand listed companies (World Federation of Exchanges, 2026) [2], while 332 million formal MSMEs carry a $5.7 trillion financing gap (IFC, 2025) [5]. Retail capital formation is already converging on revenue: 71.2% of US crowdfunding commitments in 2024 went to post-revenue issuers (SEC DERA, 2025) [106] that still cannot sell their revenue directly, only equity or notes. RSX supplies the missing instrument and the missing venue: revenue shares listed under ticker symbols; disclosure continuous and oracle-attested rather than quarterly; an SME board reporting to traditional-exchange standards, with a published graduation path to the main board. And the punchline that makes it an exchange rather than a funding portal: RSAs trade. Secondary liquidity is the endgame; a revenue-index family (sector indices of live, attested revenue streams, sliced into fractional units) is the roadmap beyond it, flagged clearly as post-liquidity.

Be precise about who this venue is for. Napkin the acquirer will only ever buy a curated fraction of the deals it sees; RSX is designed for peer-to-peer accessibility for the 99.99% of companies Napkin cannot and will not acquire but who still need liquidity. On RSX they can sell equity in part or in full, or raise entirely non-dilutively through revenue-share agreements. The founder's intent, verbatim: "We want to help, serve, and honor all." NPKN is the curated portfolio; RSX is the open venue.

The sequencing is NPKN-first, stated plainly: Napkin's portfolio companies are intended to be the inaugural listings, and $NPKN itself is intended to be an early listed asset on RSX, subject to the venue's rules like any other instrument. The operator's token earns no exemption from the operator's standards. Napkin is pledged to RSX as the first true, valuable real-world asset trading on the exchange, giving retail access to private markets "almost like an ETF would, but better": diversified private-market exposure through NPKN, single-name exposure through individual RSX listings, both on the same rails. The first believer in the exchange is the company that built it.

11.6 The Open Playbook: Seeding the Supply Side

Napkin chose the best targets so it could perform, and become the successful case study for roll-ups everywhere who want to do the same thing. NapkinDeals.com deliberately allows everyone to copy the model: the same data, the same access, the same network to capital [3]. That is not carelessness with a trade secret; it is supply-side strategy. Every acquirer that copies the playbook on the platform is a future RSX listing candidate, and the flywheel runs in one direction: the platform breeds roll-ups, roll-ups become listings, listings create liquidity, and liquidity draws more acquirers onto the platform. Napkin is not defending a secret. It is seeding its own exchange's supply side.

11.7 The Inverted Fee Curve: Entrepreneurs Ride Free at Scale

Settlement rails answer how value moves through the exchange. The question that determines who the exchange ultimately serves is different (who pays for the venue) and it is where RSX breaks hardest with precedent. Every marketplace in history has followed the same monetization arc: subsidize users to build the network, then raise the take rate on those users once they are locked in. Shopify (the best-in-class comparable for SME platform economics) runs a blended take rate of roughly 2.84% of merchant GMV, and that rate is structurally flat: a merchant pays the same effective rate whether Shopify has one hundred thousand stores or four million (Shopify annual filings, 2022: $197.3B GMV, $5.6B revenue across ~4.4M stores) [77]. The platform scales; the entrepreneur's cost does not fall.

RSX inverts this: deliberately, and as a matter of published protocol economics rather than marketing. The founder's revenue-share model [78], benchmarked against Shopify's disclosed GMV and take-rate history [77], is built on one governing rule: as the network grows, the entrepreneur's cost falls toward zero, and the funding burden migrates to the capital side: the investors and institutions who have both the greatest ability to pay and the greatest commercial benefit from the deal flow the network produces.

The illustrative curve (directional modeling, not a projection; fee parameters are set within published bounds and adjusted only through governance):

Read the last row again. At terminal scale the model still produces on the order of $2B in annual protocol revenue at a 1% blended take (comparable to Shopify's entire 2022 revenue [77]) while the entrepreneur transacts free. The network never needs to squeeze its most vulnerable participants, because the participants extracting the most value from the network's data, deal flow, and settlement rails are the ones paying for it. This is the same logic that lets a Bloomberg terminal cost $25,000 a year while the market data inside it originates from companies who pay nothing to be quoted: price the side with the profit motive, not the side creating the underlying value.

Three design commitments make the curve credible rather than a bait-and-switch:

• Published bounds, governance-throttled. Fee parameters live on-chain with a published floor and ceiling for every participant class. Moves within the bounds follow the network-scale schedule automatically; moves of the bounds themselves require governance approval through the full timelocked proposal lifecycle of Section 14. No stealth repricing.
• One-way ratchet for entrepreneurs. Merchant-side fees may fall ahead of schedule; they may never rise above the level of the prior stage. The entrepreneurs who built liquidity in the network early cannot have the terms turned against them later. The failure mode of every Web2 platform they left behind.
• First-mover economics, honestly stated. In early stages the take rate is high (5% at genesis) because early-stage networks must fund themselves and the capital side is not yet present at scale. We state this plainly rather than hiding it: the entrepreneurs who pay the most are earliest, are fewest, and receive the network's largest reciprocal rewards through early access, priority listing, and ecosystem incentives.

The endgame is a market where listing, transacting, and reaching liquidity costs a founder nothing, and the people funding the rails are the ones profiting from what founders build.

11.8 Sovereign Rails: The Stablecoin Pitch

An exchange needs settlement rails. Ours are designed for a world that has already arrived: national-currency stablecoins, licensed by the sovereigns whose currencies they carry. The pitch to any treasury is two lines long. First, settlement volume: commerce that settles in your currency's stablecoin is commerce that prices in your currency. Second, reserve demand: every unit of a national-currency stablecoin in circulation must be backed by reserves in that currency, meaning deposits and government paper. A stablecoin at scale is structural demand for your own sovereign debt.

This is not a thought experiment. The evidence tour, all real, all licensed:

• UAE: AE Coin became the first CBUAE-licensed dirham stablecoin (December 2024), and in 2026 the federal government authorized it for payment of public fees. A private issuer wired directly into sovereign treasury flows [69]. Zand followed with the first regulated multi-chain AED stablecoin on public blockchains (November 2025) [69].
• Japan: JPYC launched the first regulated yen stablecoin (October 2025) targeting ¥10 trillion in circulation (a structural JGB buyer in the making [70]) while MUFG, SMBC, and Mizuho signed a joint trust-based yen stablecoin MOU on the Progmat platform (June 2026) targeting ¥1 trillion of B2B settlement by 2028 [70].
• Singapore: XSGD has cleared over US$18B in cumulative on-chain volume and holds more than 70% of Southeast Asia's non-USD stablecoin market under MAS's single-currency framework [71].
• Hong Kong: the Stablecoins Ordinance took force August 2025; the first two licences (March 2026) went to HSBC and to Anchorpoint. A Standard Chartered joint venture with Animoca and HKT, proving crypto-native firms can sit inside a licensed sovereign-currency consortium [72].
• United States, sub-sovereign: Wyoming's FRNT launched August 2025. A US state issuing its own stablecoin with Franklin Templeton managing reserves and the reserve interest statutorily earmarked for the state school fund [73]. A US state already did this. The template exists.
• Canada. Our home lane: QCAD became the first CAD stablecoin qualified by final prospectus (November 2025) under the CSA's VRCA framework [74], with TD Bank phasing in as reserve custodian [74]; Tetra Trust's CADD carries Shopify and National Bank in its consortium [74]; and the federal Stablecoin Act (Royal Assent March 2026, Bank of Canada as supervisor) creates a statutory issuer category from 2027 [64]. The CAD stablecoin market is regulatory-ready and commercially nearly empty.
• The macro proof: GENIUS-Act-era USD stablecoins turned issuers into captive buyers of US government debt: Tether alone reported roughly $127B of Treasury exposure by mid-2025, a top-20 holder globally [75]. Stablecoins are sovereign-debt demand at scale. Every treasury that understands this will want its own lane.

We are honest about sequencing: no sovereign has yet handed its rails to an unlicensed foreign protocol, and Brazil showed that some states will tax and ban rather than partner [76]. The credible path is to become the settlement and distribution layer for existing licensed issuers first (QCAD or CADD in Canada, and their peers abroad) and earn the direct mandates later.

11.9 The Governance Destination

"It is not a DAO, it is a digital nation.": founder's working papers, 2024

The path is staged, and honest about where it starts: company, then foundation, then digital nation. The destination is a protocol governed by a constitution written in code and votable: elected representatives fused with liquid democracy, so delegation can shift every epoch; annual on-chain budgets with published audits and the power to remove those who fail them; parameter-review and data-retention windows pre-committed on an epoch cadence, so protocol hygiene runs on a calendar, not on discretion. The memberless Cayman foundation of Section 9 is the steward of that transition, and $NPKN's timelock-and-multisig governance is its embryo. Multi-jurisdiction distribution is governance risk design, not ideology: no single regime, bank, or intermediary is a single point of failure for the network.

11.10 Origins and Destinations: The Six-Year Arc

Eight Destination States

Where can RSX become? Eight destination states, in dependency order. Each one enables the next.

The first three are the load-bearing stack, and each generalizes something Napkin already runs. The settlement layer is Subsection 11.3 at terminal scale, and its pilot network exists: the portfolio's companies are the first settled revenue streams, the Proof of Balance Sheets feeds its attestation prototype, running in production before RSX exists. The listing venue inherits that book (Subsection 11.5). And the credit-data utility may quietly become the most valuable of all: continuous, oracle-verified operating data on private companies that today produce none, offered as neutral risk ratings and underwriting APIs to a lending industry that prices small business half-blind: Napkin's deal database and live portfolio feeds are that utility in miniature, pointed inward. One caveat, flagged ourselves: the talent-capital market is the most legally novel leg (hiring agreements paid in revenue share must survive employment-law review in every jurisdiction) and it is roadmap, never a launch feature.

What Six Years Did Not Change

The RSX archive spans 2020 to 2026. Plans changed constantly; these nine ideas never did, and their persistence is the best durability evidence a vision chapter can offer:

• Revenue share as the native primitive: neither debt nor equity.
• Settlement split at the revenue event itself: manual today, automatic tomorrow.
• A three-sided market: capital, companies, and talent in one venue.
• Continuous, data-driven underwriting as native infrastructure.
• Secondary liquidity as the endgame: tradeable RSAs make it an exchange.
• Fee inversion. The platform engineered against its own rent-seeking.
• Staged decentralization ending in community self-government.
• Settlement rails sovereigns actually want, in their own currencies.
• Participation over extraction, restated in every era's idiom.

Origins

The earliest documents in that archive are not written like white papers; they are written like conviction (an albatross crossing an ocean, a coming wave of liquidity) and this paper's discipline has been to translate that language into engineering, not to transcribe it. One image deserves precision. Era 5 is named Albatross (Section 3) for a reason: albatrosses cross oceans by dynamic soaring, extracting energy from the wind-speed gradient between layers of air, traveling thousands of kilometers while almost never flapping. That is the network RSX is designed to be. Its gradient runs between capital-rich participants searching for yield and commerce-rich participants searching for capital; the settlement layer extracts liquidity from that differential, and the network moves without burning intermediary fuel. The wave of liquidity was never a metaphor for hype. It was a metaphor for physics.

The roadmap itself was drawn in 2020, its five-era discipline inspired by the Cardano roadmap's named eras: a vision of peer-to-peer infinite rails where everyone has access to liquidity, reaching every continent, Antarctica included, stewarded by the foundation the founder has always called the Albatross Foundation (formally, the Napkin Foundation of Section 9). Napkin stands today at the boundary of Eras 2 and 3. And as it advances, it does not merely launch RSX: it becomes the first proof, lists on the venue it built, and pledges itself to the rails like any other company.

11.11 Genesis Intent

Founder Reserve: Disclosed Now, Not Discovered Later

The founder's RSX allocation will be set at RSX genesis with a target of no more than 15% of supply, governed by the same Liquidity Dial discipline as $NPKN: a capped allocation, sales bounded by a published dial, disclosed on-chain [22]. It will be held through a Canadian tax counsel-approved structure (CCPC or trust), with CRA income-versus-capital-gains treatment and departure-tax exposure planned for in advance and flagged here, in public, two years early. Transparent, not hidden. If that number or structure changes at genesis, the change and the reason will be published.

What We Are Not Promising

Napkin intends to launch RSX. Any RSX distribution will be retroactive, snapshot-based, free, and its parameters announced only at snapshot. Staked NPKN duration is expected to be a snapshot input; we say no more than that. No allocation formulas, no values, no yields, no dates. $NPKN must stand on portfolio cash flow alone: RSX is upside optionality, never the reason to buy.

Team & Advisors

Leadership

• Matthew Liberto: Founder & Chief Executive Officer
• Grant Anderson: Co-Founder & Executive Vice President
• Brett Robinson: Chief Financial Officer
• Josh Sturgeon: Chief Executive Officer, Napkin Labs, managing EmberTribe, Jamm Media, Podsicle, and Task Magic
• Ryan Persaud: Chief Executive Officer, NapkinDeals.com
• Asif Nasim: Chairman

Compliance is not a department here: securities counsel sits at the table for every structural decision, and the posture is compliance-first by design.

Board

At the time of writing the board comprises Asif Nasim (Chairman), Larry Mercado, Grant Anderson (Vice President), Brett Robinson (CFO), Matthew Liberto (CEO), and Suzana Seis-Manto (Secretary/Controller).

Behind the team sits the machine they built: the NapkinDeals.com marketplace, the AI workforce, and the operating history described in Section 4. Together they are proof that "AI-native acquirer" is a description, not a slogan.

Advisors

• Chris Haunschild: Rimon Law, New York; roughly $100B of transactional experience
• Andrew Dalton: CEO, Aument Capital (~$2.5B AUM)
• Nick Naamou: Director, Hedge Fund Association; CEO, Dubai Ventures
• Christopher Hughes: former CTO, Full Tilt Poker; Ethereum developer
• Estfandiar: Director of DeFi; TurtleDAO, Omniscia, W3BCloud
• Pat Hourigan: DEFYCA; formerly Bear Stearns
• Jeff Courtney: Wiser Funding
• Kilian Henner: Astro Cave
• Leonardo Voss: LVT
• Atticus Francken: Snickerdoodle Labs
• Alex: Rome Blockchain Labs
• Chris Kuriakose: Angel investor

Strategic Partners

The Stewardship Doctrine

13.1 Why Ethos Is in a Token White Paper

Because tokens do not fail on cryptography. They fail on people.

The 2024-2026 record is unambiguous. Code audits held; stewards didn't. Insider unlocks dumped into thin floats, discretionary "buybacks" timed around private information, foundations that became founder piggy banks, celebrity launches that round-tripped to zero in a news cycle. The market learned, and it learned fast: retail now prices founder conduct within hours. A screenshot of a hidden allocation moves price faster than a quarter of earnings. Conduct is a disclosure now, whether you disclose it or not.

So we treat stewardship as what it has become: a value-accrual mechanism. Every mechanism in this paper (the Engine, the NAV tenders, the attestation stack) is a promise that somebody has to keep for years. The keeping is the product. A token backed by real cash flow is only as good as the discipline of the people directing that cash flow, and discipline you announce is worth nothing. Discipline you can document is worth everything.

This section is that documentation. Nothing below was written for this paper. It is quoted from the operating system of the company: the versioned prompts and playbooks that literally run Napkin's 62-agent AI organization every day [79]. Our values are not a poster in a lobby. They are executable instructions in production infrastructure, with an audit trail. That distinction is the whole point.

13.2 The Operating Principles

The Five Values

Napkin's AI organization runs under five non-negotiable values that override any operational objective. Verbatim, from the system that approves every deal [79]:

• Integrity first. Never misrepresent a deal, a valuation, a relationship, or a capability. If a deal looks wrong, flag it: even when flagging it costs revenue.
• Serve entrepreneurs. Every decision is tested against one question: does this make an entrepreneur's life easier? If the answer is no or unclear, reconsider.
• Honor the rest. The entire organization (every agent, every pipeline) pauses for a structural weekly Shabbat rest cycle, kept regardless of operational pressure. An organization that keeps an inconvenient commitment every single week is an organization you can lend capital to.
• Proportionality. Minimum force necessary. Don't convene a board for a routine question; don't wave through a decision that deserves one.
• Transparency with founders. Leadership is never surprised. Bad news travels fastest, by design.

"If a deal looks wrong, flag it: even if flagging it costs revenue."

The values have an operational shadow: the Book's covenants. The fallow share that keeps roughly a seventh of the machine's capacity at rest, the tithe-style reserve set aside before cash is deployed, the rest closes that halt the organization on a calendar rather than a mood, and the Great Reconciliation that opens every cycle by naming failures before claiming wins: these are stewardship rendered as arithmetic. Confession precedes celebration, and the sequence is enforced by the operating system, not by memory.

The Immutable Principles

Layered on the values are seven principles the operating system marks immutable, binding at zero revenue and binding at a billion [79]. The ones that matter most to a token holder:

Integrity is non-negotiable.

Never misrepresent anything to close anything. A reputation takes years to build and one lie to destroy. When an error is found, the record is corrected and the affected parties are told: first, not last.

The mission is the mission.

A transaction that makes Napkin money but harms the entrepreneur on either side is a transaction we don't do. That filter has teeth precisely because it is expensive.

"Revenue that comes at the cost of the mission is not revenue: it's debt."

Human dignity in every transaction.

The people behind the entities in our deal graph are not data points. They are sellers exiting businesses they built with their lives, and buyers risking capital on a vision. Every seller gets a response. Every buyer gets honest information. At 10,000 acquisitions this principle gets harder, not softer, which is why it is written into the machines, not left to mood.

Compliance is architecture, not policy.

Compliance at Napkin is not a checkbox reviewed after the fact. It is built into the message bus, the database constraints, the permission boundaries, and an immutable audit log. An independent Legal Auditor reviews every deal command, and no executive (human or AI) can override a regulatory flag. When someone asks to skip compliance just this once, the answer is architecturally no.

Earn the right to scale.

Premature scaling is the most common cause of startup death, and it would be the most common cause of token death if tokens forced companies to admit it. Each stage is earned before the next is attempted. Patience is not passivity. It is discipline. This is why the roadmap in this paper runs 15 to 100 before it runs 100 to 10,000.

The Twelve Frameworks

Every material decision at Napkin runs through a documented decision stack: twelve frameworks, each tagged to the operator who proved it, each with a trigger and an action. Not philosophies to admire. Patterns that execute.

The stack has an order. Mission filter first: does it serve entrepreneurs? Competence check second: are we qualified? First principles third: what is actually true here, beyond what the market believes? Inversion fourth: what kills it? Only then does a deal reach approval, and afterward, Grove's rule decides whether it worked, and Dalio's rule decides what we learn if it didn't. Token holders are not asked to trust our judgment. They are shown the machine our judgment runs on.

13.3 The Selection Engine: Gold Miners and the Buy Box

The single greatest threat to a portfolio token is adverse selection: quietly accumulating the deals nobody else wanted. Our defense is a thesis and a filter.

The thesis is the Gold Miner Thesis: in a gold rush, sell shovels, and map the terrain. Napkin's sourcing funnel has processed on the order of 60,000 acquisition opportunities [3]. Most will never be Napkin acquisitions, and that is the design. The overwhelming majority of deals don't fit our buy box, but the data about them is valuable to everyone: it becomes marketplace inventory, buyer-matching intelligence, and live transaction-intent data. Revenue and signal, never balance-sheet risk. The funnel monetizes what it rejects. That inverts the usual roll-up incentive to lower the bar when deal flow slows.

The filter is the buy box. A documented qualification gate every deal must fully pass:

• Real business: operating and revenue-generating, not idea-stage
• Realistic price: inside a defensible valuation range for the industry
• Transferable: runs without the founder, or has a credible succession plan
• No red flags: no pending litigation, regulatory exposure, or fraud indicators
• Disciplined fees: a fixed 8-12 percent band the system itself enforces; exceptions require founder sign-off, and the architecture holds the deal until they do

Sector-specific screens stack on top: recurring-revenue quality and customer concentration for software, regulatory screens before sourcing for financial businesses, payer concentration for healthcare. Escalation is proportional to size, up to mandatory advisory-board review and founder approval for the largest transactions. And the funnel is not hypothetical: at the time of writing, 60,000+ opportunities filtered against exactly this buy box (tangible assets, durable cash flow, LBO-serviceable economics) have produced the four signed construction transactions of Section 4 [3]. The filter holds under load.

Around the filter sits a trust layer: trust before transaction. The Verified Buyer identity (vetted capital, vetted intent, a badge sellers can rely on) filters tire-kickers out of the marketplace, which keeps the data honest, which keeps the funnel honest, which keeps the portfolio honest. Every layer of the machine is selecting for the same thing.

For NPKN holders, this is the anti-adverse-selection engine protecting NAV. The portfolio behind your token is not what showed up. It is what survived.

13.4 The Record

Name the ambition plainly. The most successful serial acquirer in history, Constellation Software, has acquired more than 1,000 companies over roughly three decades (Constellation Software filings, 2025) [25]. The largest agency consortiums (Omnicom's network among them) sit in the low thousands of accumulated entities (Omnicom Group reporting, 2025) [80]. Napkin targets 10,000. If achieved, it is the largest roll-up in history. There is no modest way to say that, so we won't pretend to.

Here is what makes it thinkable rather than delusional: the bar is lower than people assume. The record holder averaged around 30 acquisitions a year for most of its run, constrained by human diligence, human integration, and human reporting [25]. Those are precisely the constraints Napkin's architecture removes: AI-driven sourcing and qualification at funnel scale, standardized deal terms, and the automated onboarding described below. The record was set by hand. We intend to beat it with software.

Two things distinguish the attempt beyond size. First, fairness: founders who sell into the Napkin portfolio receive common stock on equal footing. No preferred-over-common games, no liquidation stack engineered to strip the operator who built the business. And this is already company policy, not a token-paper promise: per the company's own shareholder reporting, the committed long-term structure is a single class of common stock. No preferred, disciplined on dilution, with buybacks of the earliest investors over time [3]. The cap table practices what it preaches: the founders hold 15 million shares each, exactly equal, alongside roughly $19.8M CAD of outside capital across roughly 78 shareholders, and no preferred stock exists [3]. Single-class common was the committed destination through years of shareholder reporting, and the Arrangement of Section 6.7 reaches it: the moment the plan of arrangement takes effect, every legacy class and instrument collapses into one class of tokenized common shares, and the destination becomes the register. And through the structures in this paper, that ownership has transparent, real liquidity instead of a decade of lockup and hope. Roll-ups have historically extracted from founders. This one recruits them as aligned holders, and 10,000 aligned founders is a distribution moat no marketing budget buys.

Second, honest sequencing. Fifteen acquisitions have been completed since 2021 as this paper goes to print, a count that includes the divestitures that pruned the portfolio to its current consolidated core, an honest scoreboard most acquirers never publish. We do not ask anyone to underwrite 10,000 on that base. The next milestone is 100, and we publish per-deal economics as we go, so by the 100th deal the market can judge the machine on attested numbers, not narrative. Then the record conversation gets serious. Claims scale with evidence. That, too, is stewardship.

And the candor is not a launch-day costume. Napkin's own shareholder reporting contains a section literally titled "Candor: Deals That Did Not Close", naming the transactions that fell through, and why [3]. Most issuers publish their wins. We publish the scoreboard, to our own shareholders, before any token holder existed to impress. That is the disclosure ethic the token program inherits.

Stewardship in action: the Nobody Loses Covenant

The Conversion of Section 6.7 is where the doctrine stops being prose and starts costing money. Before the token era opens, every legacy shareholder is offered a voluntary cash exit at or above invested cash basis, priced tranche by tranche under the company's confidential buyback model, a company document prepared for the board and counsel [3]. The tranche prices stay confidential because they are individual shareholders' personal economics, not marketing material; the rules that govern them are published in full. Those rules carry a name, the Nobody Loses Covenant, and three terms: a Cash Floor (no shareholder is ever cashed out below hard cash invested per share), a Universal Roll (every holder may exchange paper for paper at ratios that preserve proportional ownership, so upside is never confiscated), and No Forced Cash (tag-along and drag-along mechanics only ever drag a holder into the roll, never into cash below basis). Behind the covenant stands the Founder's Backstop: any holder whose contractual exit falls below invested basis is made whole personally by the founder, from his own assets, not the company's.

Read those terms as a stewardship exhibit, because that is what they are. The shareholders who carried the early risk are honored with a cash win before a single token exists; their exit is the last act of the old covenant, kept in full. The founder collateralizes the floor personally, which is the difference between a value on a poster and a value with recourse behind it. And the whole mechanism is disclosed in advance, here and in the arrangement materials, so anyone can audit the conduct against the covenant. Section 6.7 carries these as terms; this section claims them as character.

13.5 Zero-Trust Onboarding at Scale

The standard objection: nobody can govern 10,000 companies. The honest answer: nobody can govern 10,000 companies with the tools public markets use today. Quarterly reporting is a 90-day-old photograph, assembled by hand, sampled by auditors, published on a lag. That machinery barely governs 500 companies. It was never going to govern ours.

So we replace it with something more rigorous, not less. Every acquired company is onboarded onto a blockchain-native reporting spine from day one:

• API-fed financials: banking, payments, and accounting feeds wired directly into the consolidation layer. Revenue observed at the source, not transcribed at quarter-end.
• The attestation stack: a single audited IFRS consolidation, quarterly examination-level attestation of per-deal financials, and real-time oracle-published NAV. The full Proof of Balance Sheets architecture of Section 8, applied to every new company at close.
• Zero-knowledge direction: attestations that prove revenue claims (thresholds met, covenants satisfied) without exposing raw books, so verification scales without leaking commercially sensitive data.

Compare that honestly with a traditional public onboarding: months of bank syndication, sampled audits, and standardized disclosure produced by humans on deadlines. Our pipeline verifies more, continuously, automatically. Zero-trust onboarding is the breakthrough that turns 10,000 from a slogan into an operations plan: the marginal cost of governing the next company falls with each one, because the governance is code.

13.6 The Organizational Model: Profit With Purpose

Structure is where stewardship either becomes real or becomes theater. We are copying the two structures with the longest track records in this industry, and naming them plainly: IOHK and Cardano, Ripple and the XRP Ledger. An engineering organization builds; the protocol belongs to its ecosystem.

Napkin Labs is the protocol's developer organization. A research-driven engineering house in the mold of IOHK's peer-review culture, where design decisions are published, argued, and defended before they ship, not announced from a stage. We consider that culture a form of intellectual-capital stewardship, and the explicit alternative to the celebrity-founder model this industry keeps rewarding and then burying.

The Napkin Foundation (Cayman, memberless, orphaned) stewards the ecosystem treasury. Before describing it, one sentence with no ambiguity: we are not a nonprofit and we will not pretend to be one. We are building the most profitable machine we can, and hard-coding a slice of it to widen the market we profit from. The model here is Andrew Kuper's LeapFrog Investments, the profit-with-purpose pioneer that proved serving underserved markets is not charity adjacent to the business. It IS the growth market, an underwriting and returns edge disciplined enough to attract institutions like Temasek and AIA at multi-billion-dollar AUM (LeapFrog Investments) [81]. Purpose, done honestly, compounds returns. Purpose done as theater destroys them. We know which one we are buying.

So the Foundation is not a charity fig leaf. It is the ecosystem's R&D and market-development engine: grants and bursaries for open attestation and privacy tooling, builder education, and education for the SME founders who are this network's sellers, buyers, and future holders. Every dollar of it is commercially selfish at one remove. A bigger, healthier on-chain SME ecosystem directly grows the portfolio's deal flow and the token's addressable market.

The commitment is predetermined, not discretionary. A published Purpose Allocation targets 1 percent of Token Cash Engine inflows routed to the Foundation's social-capital arm: fixed, hard-coded, auditable on-chain like every other flow in this paper, and subject to final legal and covenant review. It is never senior to bond obligations, never senior to holder economics. The percentage is set before the money arrives (the tithe discipline of the founder's faith, applied to a treasury) so that giving is a standing commitment, never a marketing decision. Separately and personally, the founder intends to pledge a share of his own token distributions to the Foundation; his conviction, his tokens, not a levy on holders.

Over time the Foundation's arm will deploy mission-aligned lending tools: revenue-based financing for SMEs in underserved markets. LeapFrog logic again: these loans are designed to be profitable and to widen the funnel of businesses that grow into acquisition targets. Not giveaways. A flywheel.

The founder's own framing is the mandate: NPKN should be "freedom for people. Not so centralized that the greed of a few developers and large holders kills the value for everyone." That is not a vibe; it is a design constraint, enforced by commitments a reader can check. No celebrity-founder cult: the cap table is published, the founder allocation is capped at 20% of supply, and every founder sale runs through the published Liquidity Dial, disclosed on-chain within days. No hidden hands: allocations, unlocks, and treasury releases are mechanical and on-chain. No developer capture: the protocol's economic engine is hard-coded and non-discretionary, and the adversarial-design governance framework of Section 14 exists precisely to hold these commitments when they get expensive. This section states the values; Section 14 carries the mechanics that enforce them.

13.7 The Moat Nobody Can Fork

Everything else in this paper can be copied. The contracts are inspectable, the structure is described, the playbooks will leak. What cannot be copied is a decade of flagged deals that cost us revenue, founders treated with dignity on the worst day of their professional lives, audits published on time when the numbers were ugly, and a compliance architecture that refused a shortcut every single time someone asked. Stewardship compounds like the portfolio does (one kept promise at a time) and it is the only asset here with no fork button.

"The system won't let you. That's by design."

Adversarial Design, Protocol Integrity & Governance

Most white papers hide their attack surface. They describe the mechanism that works and stay quiet about the twelve ways it breaks. We do the opposite, here, at length: because a protocol whose entire pitch is transparency has no business being opaque about how it could be attacked. If we ask you to trust attested numbers over promises, we owe you the adversary's view of those numbers first.

So this is the red-team report on $NPKN. It assumes a well-capitalized, patient, technically fluent adversary (a rival protocol, a short fund, a governance raider, a compromised insider, a bridge crew) who has read every prior section and is hunting for the seam. We name attacks with real precedents and dollar figures, and every one ends with an honest sentence about the risk we could not engineer away. A red-team report with no residual risk is not a red-team report; it is marketing.

Three principles govern the design. Assume adversaries: every mechanism is specified against a hostile counterparty who knows the rules, not a cooperative one who follows the spirit. The published buyback wallet, the quarterly tender, and the audit lag are all targets, built expecting attack on day one. Prefer mechanisms over promises: wherever a discretionary decision could become a hard-coded rule or a covenant with a penalty attached, it did: because the market's verdict is settled that accrual depending on a future decision is priced at zero [34]. Make honesty the cheapest strategy: we gate step-ups on attested cash flow because cash is expensive to fake, invert the disclosure penalty so our own silence costs us, and embrace the security wrapper because the law that makes fraud expensive is the law that makes our ownership claim true. The move throughout is to price dishonesty above honesty and let self-interest enforce the rest.

The Threat Model at a Glance

Sixteen threat classes spanning every domain, with the mechanism that does the most work against each. The subsections treat each in full: vector, precedent, mitigation, residual risk.

Liquidity Warfare & Vampire Attacks

The oldest fear a token founder carries is the vampire: a rival who forks the code, bolts on an incentive, and drains the liquidity in a weekend. The real history is more instructive than the fear.

The SushiSwap autopsy

In September 2020, SushiSwap forked Uniswap v2, added a token, and paid it to anyone who staked Uniswap LP positions. It worked spectacularly for about two weeks: on September 9, 2020, the migration pulled roughly $810M (about 55% of Uniswap's liquidity) in a single day, with over $1B staked within ten [28]. Then the decade. By 2025-2026 SushiSwap's TVL sat near $102M, down 98.7% from its ~$8B peak, on roughly $4.5M annual revenue through chronic leadership churn, while Uniswap held over $4B in TVL and collected roughly $985M in protocol fees from January to October 2025 alone [28]. A 2025 Journal of Operations Management study [53] found vampire attacks routinely strengthen incumbents by forcing them to ship faster. Forks copy code. They do not copy liquidity network effects, audit history, integrations, brand, or teams.

For $NPKN the threat is weaker still, because the moat is not code. You cannot fork a legal claim on real audited operating companies, a EUR 500M Vienna-listed bond program, and a consolidated audit. A forker owns a token pointing at nothing. No share of NewCo, no title to the portfolio or its cash flow, no acquisitions to consolidate. The precedent is Figure on Provenance: anyone could fork the code, but Figure's $17-20B tokenized loan book never moved, because the moat was origination and legal ownership [54]. Our BSL 1.1 license buys a multi-year head start, but we hold no illusions: licenses do not prevent forks, they schedule them. Residual risk: a fork cannot take our assets, but a well-funded copycat can still fragment speculative volume and confuse retail short-term.

Hardening the Engine against MEV

Transparency's corollary is that a published, pre-committed, on-chain buyback is a searcher's dream: public wallet, formula, and schedule, so every slice can be sandwiched (front-run the buy, sell into the lift, pocket the spread from the holders the burn should benefit). Sandwich bots have extracted hundreds of millions from predictable flow [82]. We harden the Engine on four fronts:

• Private orderflow. Buyback slices route through encrypted private channels (Flashbots Protect / MEV-Share relays), not the public mempool, so a searcher cannot see the buy or wrap it.
• Batch-auction execution. Where supported, buybacks clear through uniform-price batch auctions (the CoW Protocol pattern), denying sandwichers the deterministic ordering their attack needs.
• Randomized slicing. Slice sizes and intra-window timing are randomized within published bounds, so a public formula does not collapse into a predictable next trade.
• Commit-reveal tenders. The quarterly Dutch auction uses hashed bids revealed after close, so the clearing price cannot be front-run or last-look sniped.

Residual risk: private orderflow and batch auctions cut extractable value sharply but not to zero. A determined adversary with builder relationships keeps a residual edge on the margin.

The wash-down to tender-snipe combination

Here is a combination a single-domain analysis misses, chaining market manipulation into tender arbitrage. The tender executes at the better of market price or discount-bounded attested NAV, and that floor is the vulnerability: an adversary who pushes the market price down (wash trades, spoofed depth, coordinated selling into a thin book ahead of the window) can accumulate cheaply and tender back to the Engine at the NAV floor, extracting the spread from every other holder's cash flow. Each leg has precedent: wash-trading-as-a-service is now a prosecuted crime [59]; tender-capture is the weaponized cousin of Saba Capital's 2023-2024 campaigns forcing closed-end-fund tenders to harvest NAV discounts [40]; and pushing a thin-market price to inflate a mechanical claim is how Avraham Eisenberg drained roughly $110M from Mango Markets [83]. We defend at the seam: the tender prices against last-attested NAV with a notice period (not spot), is capped per quarter, clears pro-rata via commit-reveal, and (decisively) $NPKN is an ERC-3643 permissioned security, so the wash and vampire legs cannot run through anonymous wallets. Residual risk: within the whitelisted holder set, coordinated selling into the pre-attestation window is still possible, and the tender bounds rather than abolishes a well-timed discount's profitability.

Reflexivity between the buyback and the tender

Both legs draw from the same Engine flow (the 50/50 split of a fixed FCF percentage) which creates a reflexive interaction under stress. In a deep-discount quarter the buyback buys and the tender opens at the NAV floor, both consuming one fixed pot; if a few large holders rush the tender, the quarterly allocation can be exhausted before retail reaches it. A bank run on a capped tender. Worse, each burn raises NAV-per-token and thus next quarter's tender floor, so the fixed cash buys fewer tokens even as demand to tender rises. The canonical reflexive-NAV blowout is the Grayscale GBTC discount reaching roughly -49% in December 2022, whose reversal helped detonate Three Arrows Capital (over $10B) and the 2022 contagion [40]; the canonical death spiral is LUNA/UST erasing roughly $40B in days [20]. Our defenses are rationing and honesty, not a claim to abolish physics: the tender is pro-rata within the cap so no one drains the queue, we say plainly the mechanism bounds the discount and no more, and the no-issuance-below-NAV covenant prevents the Terra reverse leg. Residual risk: in a persistent multi-quarter discount, tender demand can outrun the fixed cash that funds it and the reflexive dynamic between the legs is real. We ration it, we do not eliminate it.

Cross-Domain Combinations, Contagion & the Bridge Surface

The attacks that end protocols cross domains. A technical bug becomes an economic event, one company's fraud becomes the whole system's freeze. This is the part a domain-by-domain analysis is least equipped to see.

One failed company, the whole NAV

$NPKN's value derives from a single consolidated perimeter. The right accounting design, but it means one acquired company can become a systemic event. If one opco turns out to be a RealT in miniature (cooked books, revenue on deals that never closed, an undisclosed liability), it does not merely zero its own deal. It can force a restatement of the consolidated accounts, trigger a qualification or withdrawal of the quarterly ISAE 3000 attestation for the entire perimeter (tripping the oracle's staleness breaker and freezing the Engine for all holders) and, if the writedown is large enough, breach the bond's coverage test and auto-suspend the token tranche. The precedents are holding-company graveyards: Steinhoff lost roughly €10B, about 96% of equity, when one fraud surfaced [84]; Abraaj, a $14B PE firm, collapsed when commingling in one fund detonated the house [84]; NMC Health imploded on roughly $4B of hidden debt [84]. We mitigate with component-materiality caps on any subsidiary's weight, ring-fencing with no cross-guarantees so one default cannot cross-default the rest, and audit scope over the AI-operated accounting stack. Residual risk: a consolidated portfolio is a correlated instrument; a large-enough single-company fraud can restate the NAV, freeze the mechanism, and breach the bond at once, and diversification dampens this only as the portfolio grows.

The early-stage concentration honest paragraph

The line is "10,000 balance sheets, one token." The near-term reality, at the time of writing, is a consolidated operating portfolio still counted in single digits (anchored by a Day-1 construction platform whose largest customer relationship is itself a named concentration) where a single name can be 15-30% or more of consolidated free cash flow, which makes the diversification that renders contagion survivable aspirational, not yet operational. The premise that no one deal can sink the portfolio is a property $NPKN earns toward the 100-deal milestone, not one it holds at TGE. Residual risk: for the first several years $NPKN carries real single-name concentration risk, and no mechanism changes that until the deal count grows.

The bridge as the worst surface in crypto

Bridges are empirically where crypto loses the most: cumulatively over $2.8B, roughly 40% of all DeFi exploit losses [85]. Ronin ($625M, 2022), Wormhole (an unbacked mint of 120k wETH, ~$326M, 2022), Nomad ($190M, 2022) [85]. $NPKN touches this surface three times: the Base spoke (burn-and-mint via ERC-7802 with Chainlink CCIP), the XRPL spoke, and the future RSX Avalanche L1's ICM gateway. The specific nightmare is a supply-conservation break: a bug that mints $NPKN on a spoke without burning it on the hub is an infinite-mint of a security, corrupting the per-token denominator that every NAV, buyback, and tender calculation depends on. A technical exploit that becomes economic dilution and unauthorized issuance at once, exactly the BNB Bridge failure that forged a proof to mint 2M BNB (~$586M, 2022) [85]. We shrink the surface rather than decorate it: no lock-and-mint honeypot (burn-and-mint nets spoke supply to zero, so there is no pool to drain), ERC-7802 in the token with CCIP a swappable authorized caller, formal verification (Certora / Halmos) proving cross-chain supply conservation as an invariant, and rate-capped CCIP lanes through the guarded launch. Residual risk: CCIP is itself a trusted network, the RSX gateway is a second bridge yet to be built, and bridges remain the industry's highest-loss category. We cut the blast radius and probability but cannot eliminate a cross-chain messaging compromise.

Oracle-signer compromise and attestation-lag arbitrage

The oracle transmits; it does not verify: two adversarial corollaries follow. First, whoever signs the attested NAV is a point of trust; a compromised or coerced signer could push a false NAV the Engine executes against. The nearest precedent is Synthetix's 2019 incident, where one mispriced feed let a bot mint roughly $1B notional of sETH before it was unwound [83]. We mitigate with multi-party signing, rotatable keys, deviation bounds, and a staleness breaker that freezes NAV-dependent actions while leaving transfers untouched. Residual risk: named-firm accountability is a legal deterrent, not a cryptographic guarantee. Second, the real-time dashboard is a double-edged sword: it publishes live operating data, but the NAV that prices the tender is up to 45 days stale, and that gap is tradeable. The tokenized-asset version of the mutual-fund late-trading scandal that cost billions in settlements [83]. We pin tender pricing to attested dates, add notice periods, and treat NAV-relevant data as MAR inside information. Residual risk: transparency creates a faster information surface than the attestation that governs payouts, and that lag can never be fully closed.

Counterparty and operator concentration

The stated philosophy is multi-jurisdiction redundancy, but the operator layer is not yet redundant: a single US transfer agent and ATS (Securitize), a primary named EEA venue (21X), one independent buyback agent, and (worth flagging) Chainlink for both the CCIP bridge and the NAV oracle, so a single-vendor incident could touch pricing and settlement at once. A trusted intermediary's failure cascades: FTX left roughly $8B short [83]; Multichain froze roughly $1.5B when its operator vanished [83]. Residual risk: in the early phase several mechanism operators are single points of failure, and the redundancy the corporate structure promises is not yet built at the operational layer.

Time-based attacks: cliffs, snapshots, blackout windows

Predictable calendar events are attackable even with a smooth curve. Unlock front-running: the published unlock calendar and first emissions are known dates desks short. The "unlock trade" that hit TIA (-90% after its 176M cliff) and ONDO (-87% on annual cliffs) [30]; our fixed-6th monthly-linear schedule flattens the steps but cannot erase the calendar. RSX snapshot farming: because staking duration is expected to be a snapshot input, an actor who guesses the window could stake just before, capture weight, and dump after. The mercenary pattern that broke EigenLayer's social contract and triggered roughly 150k ETH of withdrawals overnight [67]; the defense is retroactive, surprise-timed snapshots with parameters set at snapshot, plus duration weighting. Blackout front-running: whoever knows the buyback is paused around a disclosure can sell into the absent bid and buy back after: randomized slicing and TWAP resumption add noise. Residual risk: any deterministic input can be farmed by whoever learns the timing, which makes snapshot-date confidentiality itself a MAR problem, handled by insider lists, never reduced to zero.

Economic Gaming, Sybil & Oracle Integrity

Where the last subsection chained domains, this one attacks the incentive surfaces directly. The places someone games a rule for private profit at the protocol's expense.

Deal Scout self-dealing and Sybil farming

Deal Scout pays 0.5-1.5% of closed enterprise value, in vested NPKN, for sourced targets. Two attacks follow. Self-dealing: a scout submits a secretly-related seller and inflates the EV to inflate the bounty. A kickback dressed as a referral. Sybil farming: fake accounts submit recycled targets to farm points, an endemic pattern: LayerZero's own purge found roughly 800k+ Sybil addresses [67]. We pay only on closed deals that consolidate, at auditor-verified arm's-length EV, to KYC'd scouts, in vested tokens no farmer can dump, and because holding $NPKN requires an identity claim, anonymous-Sybil economics largely do not apply. Residual risk: self-dealing detection depends on diligence quality, and a scout colluding with an arm's-length-looking seller is hard to catch ex ante.

The incentive divergence nobody prices

A subtler risk lives between two correct mechanisms: scouts are paid on enterprise value, but the Engine steps up on attested free cash flow, and those point in different directions. A scout maximizing bounty wants big deals, even low-FCF ones, biasing sourcing toward exactly the deals that dilute the token's metric. Residual risk: this principal-agent divergence closes with no single rule; it must be monitored as FCF per euro of EV acquired, disclosed, and corrected by adjusting bounty terms if it appears.

Milestone gaming

"Unlock per N acquisitions" is trivially gameable. A hundred one-euro shell purchases "close a hundred acquisitions." The design refuses it: step-ups (20% to 30% to 40%) trigger on cumulative attested consolidated FCF, examined quarterly under ISAE 3000, with Qualifying thresholds (minimum trailing EBITDA, arm's-length pricing, two quarters inside the perimeter). Cash is harder to fake than transactions. Residual risk: cash-flow gating shifts the attack from fabricating deal counts to fabricating the underlying financials, which is why the verification stack, not the milestone rule, is the real defense.

RSX snapshot Sybil

A progressive, whale-capped RSX distribution invites the mirror-image of concentration: a whale splitting holdings across many wallets to farm the small-holder bonus, the way Optimism, Arbitrum, and Hop all shipped with detected Sybil clusters [67]. Here the permissioned architecture is a structural defense the industry cannot match. One KYC'd identity cannot cheaply spin up 500 wallets, because holding requires ONCHAINID claims tied to a verified person. Residual risk: identity farming (renting KYC'd identities or spreading holdings across genuine associates) raises the cost of Sybil without eliminating it.

NAV and oracle manipulation

Beyond signer compromise, the classic on-chain attack manipulates a thin market a mechanism reads as truth. The Mango pattern, roughly $110M from moving a price on a shallow venue [83]. $NPKN's mechanisms read attested NAV, not spot, publishing only auditor-signed figures with the report hashed alongside, so there is no thin spot price for a payout to trust. Residual risk: the mechanism is only as honest as the attested input, returning the exposure to auditor error or fraud in the underlying companies.

Wash trading and tender-clearing manipulation

Two market-integrity attacks close the domain. Wash trading to paint volume or price meets a contractual ban in every MM agreement, named market makers, retainer-only contracts, and monthly liquidity reporting, and it is now prosecuted [59]. Tender-clearing manipulation, shading bids or last-look sniping a Dutch auction: is removed by the commit-reveal flow. Residual risk: quant desks reconstruct MM inventories from on-chain flow, and no disclosure regime makes manipulation impossible. Only detectable and contractually breaching.

Governance Capture & Insider Abuse

A protocol that hands its treasury to a token vote hands an attacker a target. $NPKN's answer is unusual: the highest-value governance attack is neutered by design, because the Engine is not a thing anyone can vote on.

Whale and flash-loan governance attacks

The signature heist is the flash-loan vote (borrow tokens for one block, pass a malicious proposal, drain the treasury, repay) which cost Beanstalk roughly $182M in a single transaction [83] and let a hostile majority take over Build Finance DAO outright [83]. It requires a treasury controlled by a token vote. $NPKN has none: the Engine percentages, waterfall order, bond seniority, and NAV discipline are covenant-grade and non-votable, so there is nothing to flash-borrow into. The residual surface (bounded periphery parameters) is defended in depth by a 48-hour minimum timelock (seven days for oracle or compliance modules), a multisig execution gate, and an independent board veto. Residual risk: a whale accumulating a genuine long position still concentrates influence over in-scope parameters, and social-engineering a legitimate vote is slower but not impossible.

Multisig and signer compromise

The most damaging recent attacks broke signers, not contracts: Bybit lost roughly $1.5B in February 2025 to a blind-signing attack on a Safe multisig [83], and Radiant Capital lost roughly $50M in October 2024 to compromised signer devices [83]. Directly relevant, because module swaps and emergency pause sit behind Safe multisigs and Napkin runs 62 AI employees. We mitigate with hardware-isolated signing, transaction simulation before signing (defeating blind-signing), threshold requirements, and a hard rule that no autonomous AI agent ever holds a complete key. Every agent action sits behind the Fireblocks policy engine and Safe thresholds. Residual risk: multisig security depends on the operational security of human signers and their devices. A people problem no contract solves.

Insider dealing under MAR

The Vienna bond listing already subjects Napkin to the EU Market Abuse Regulation, and a serial acquirer continuously generates inside information: trading ahead of an accretive acquisition, a bad attestation, or an Engine step-up whose FCF threshold insiders see approaching. Crypto insider dealing is live and prosecuted. The first case, against a Coinbase lister, brought DOJ and SEC charges [59]. We mitigate with MAR insider lists, blackout windows, a named human compliance officer (never an AI agent), step-ups as mechanical hard-currency thresholds rather than announced-then-traded events, and a non-discretionary buyback so the issuer cannot time its own tape. Residual risk: insider trading by an individual is a perennial risk that procedure deters and enforcement punishes but no design eliminates.

Conversion execution as a threat surface

The Conversion of Section 6.7 is a chain of approvals that no adversary needs to attack for it to fail on its own: a court can withhold the order, dissenters can litigate, bondholders can withhold change-of-control consent on the EUR 500M program, tax counsel can decline to opine. Any one failure delays the token, and the design makes that delay automatic rather than negotiable: Gate G1 of Section 10 holds TGE until the court order and the consent exist, so a slipped approval slips the launch instead of shipping an unconverted token. Residual risk: the gates convert execution failure into delay, not holder loss, but the delay is real and its duration is not ours to set.

Founder-sale optics

A founder selling any shares into the Phase 1 tender can be read, uncharitably, as an insider exit dressed as a reorganization. The defense is that nothing about it is discoverable, because all of it is pre-capped and pre-disclosed: the founder allocation is bounded at 20% of supply (future founder recruits are granted from inside that pool, never from the community), every sale runs through the published Founder Liquidity Dial (up to 10% of remaining holdings per rolling 12 months, executed outside blackout windows, disclosed on-chain within days), Phase 1 tranche terms are identical to every other shareholder's, and the final split appears in the token documents before TGE. This is a stronger defense than a padlock, because a padlock is trusted until it opens while the dial is checkable against the chain every quarter; the published cohort model shows founders selling the maximum dial every year still end near 9.5% of supply while the community holds above 62%. Residual risk: optics are not controlled by disclosure, only answered by it; a bad-faith reading will circulate regardless, and the published numbers are the reply.

Market-maker abuse and bad-actor detection

The quiet way clean tokenomics gets dirty is the MM loan-plus-call-option, whose mispriced strikes create hidden sell pressure and phantom float. We use retainer-model contracts only, name our market makers, count any loaned tokens as circulating, and publish monthly liquidity reports. The strictest 2026 norm, voluntarily. Bad-actor detection runs on Hypernative and Tenderly with a rehearsed auto-pause playbook, so an exploit in progress meets a circuit, not a committee. Residual risk: monitoring catches known patterns fast and novel ones slowly, and the window between a zero-day and its detection is never zero.

The NPKN Governance Framework

Governance is where these threats are contained or unleashed. We designed it on one premise: the most important governance decision is what governance is not allowed to decide. A security's issuer obligations cannot be delegated to an anonymous vote, and the mechanisms that make $NPKN honest must be beyond the reach of the majority that might, in a bad moment, vote to switch them off.

The constitution: principles no vote can cross

These are immutable: written into the token terms and the immutable core contract, with no quorum, supermajority, or emergency path that alters them:

• The Engine exists and is non-discretionary; its percentages step only on attested-FCF milestones and can never be voted up, down, or off.
• Bond obligations rank senior to the token; the waterfall order is fixed.
• No new $NPKN issues below attested NAV per token.
• Only auditor-attested figures drive the Engine; self-reported marks never touch it.
• Locked and unvested tokens cannot stake and carry no RSX-snapshot weight.
• The disclosure covenants and their penalty inversion stand; issuer silence costs the issuer.
• The core token contract is immutable; the asset you hold cannot be rewritten.
• Sources and uses stay segregated; holder money is never the collateral behind holder claims.

The three-body model

Power splits across three bodies with non-overlapping mandates, so none can both propose and unilaterally execute a consequential change.

The Oversight & Audit Board (OAB) is the innovation: an independent body (not Napkin employees) that reviews every proposal for legality (MAR, prospectus, securities law) and constitutional scope, holds a veto exercisable inside the timelock, and oversees attestation integrity, giving holders an independent check on the numbers the Engine trusts. The Napkin Foundation stewards the future protocol and holds emergency keys but is walled off absolutely: $NPKN lives entirely inside the regulated issuer stack, never the Foundation.

On-chain versus off-chain scope

The line is drawn by law, not convenience. On-chain, token-votable (behind timelock): bounded periphery parameters: oracle-adapter provider from an approved set, compliance-module jurisdiction updates, ecosystem-reserve allocation within the ten-year cap. Off-chain, not token-votable: every issuer obligation of a regulated security (prospectus terms, M&A decisions, corporate actions, the audit relationship) which sit with the boards and companies law, because a prospectus-approved security cannot lawfully hand its issuer duties to a pseudonymous majority.

The proposal lifecycle

Every in-scope change runs the full gauntlet; the guardrails are floors set in the token terms, and sensitive parameter classes carry higher quorum and supermajority requirements.

Floors are fixed in the token terms: ordinary proposals require a minimum quorum of staked supply, sensitive classes (oracle adapter, compliance module) require both a supermajority and a higher quorum, and anything constitution-adjacent is out of scope and cannot be tabled at all.

Emergency powers and their hard limits

An exploit does not wait 48 hours, so a security-council Safe can pause without timelock, and that power is deliberately small and boxed on every side. The pause can freeze NAV-dependent actions (buybacks, tender pricing, primary issuance) to halt an exploit. It cannot mint tokens, move holder funds, alter the Engine percentages or waterfall, or touch the constitution. It never freezes transfers of already-held tokens: holders can always exit, whatever the machine is doing. Every emergency action is event-logged and disclosed, and the pause auto-sunsets and requires ratification within a stated short window, so "emergency" cannot quietly become "permanent."

The progressive-decentralization roadmap

We say the quiet part plainly: today this is centralized-but-transparent, which is the correct posture for a regulated security at launch. The issuer and boards run the protocol; everything they do is published, attested, and covenant-bound. Governance decentralizes progressively as the protocol matures and legal clarity increases: in-scope parameters hand to token vote in stages, the OAB's independence deepens, and the RSX era brings the Foundation's stewardship online. But the constitution never decentralizes: the Engine, waterfall, NAV discipline, and disclosure covenants are fixed at the start so that decentralizing governance can never mean decentralizing away the guarantees. A project claiming full decentralization on day one is either lying or unregulated; we are neither. Residual risk: progressive decentralization is a promise about the future, and any such promise is a governance risk, which is why the guarantees that matter most are the ones we made un-votable from the beginning.

Residual Risks We Cannot Fully Eliminate

Everything above ends in a residual sentence. Four risks, though, are not attack-and-mitigation pairs. They are structural facts about what $NPKN is, and pretending they are solved would be the reddest flag in this document.

Market beta

$NPKN trades in crypto markets and will fall when they fall. The Engine changes the per-token economics of a growing portfolio; it is not, cannot be, and is not designed to be price defense. Uniswap burned $596M of UNI and hit an all-time cycle low of $2.90 two months later [35]; Pump.fun spent over $350M on buybacks and sat 81% below its peak [43]; Hyperliquid, the best-executed buyback in crypto history, drew down through the H1 2026 bear despite roughly $65M a month of buying [31]. No mechanism outruns beta.

Black-swan operator fraud

The verification stack is built to catch the RealT failure mode: tokens sold on assets never owned, dividends on empty properties, a $2.72M fraud across 39 homes [58]. But a sufficiently sophisticated fraud inside an acquired company, or collusion reaching the attestation itself, is a tail we make expensive and detectable, not impossible. Named-firm auditor accountability is a deterrent backed by legal liability, not a guarantee no one ever lies successfully.

Regulatory reversal

A five-jurisdiction structure has five regulators, and securities regimes shift. We built prospectus-first because BaFin's Ethena action showed what "launch then negotiate" costs. A public-offer ban, frozen reserves, a EUR 600,000 fine, a supervised 42-day wind-down [47]. But an adverse reinterpretation in a major market could still force geo-restriction of distribution, and we premise no plan on pending legislation. The strategy reduces regulatory risk; it does not abolish a regulator's discretion.

Key-person risk

The founder and a small leadership team drive strategy. The Engine and vesting are contractual and non-discretionary (they run whether or not any individual shows up) and governance sits behind a multisig, timelock, and independent board. But an AI-native serial acquirer is, at this stage, inseparable from the people who built it, and their loss would be a real shock the mechanisms cushion but do not neutralize.

A protocol that claims to have engineered away market beta, fraud, regulators, and its own founders is not a safer protocol. It is a less honest one.

Integrity is not a disclaimer we bury on the last page. It is the feature. The one mechanism every other mechanism in this document depends on, and the only one we could not write in code, so we wrote it in incentives instead.

Risk Factors

We wrote this section to be read, not skimmed. Each risk is real, each mitigation is specific, and none of the mitigations is a guarantee. Where a risk's attack mechanics are treated at length in the adversarial-design analysis, we give the one-line version here and cross-reference Section 14. The category stays, the duplication goes.

NAV discount risk: the #1 structural risk.

A token representing portfolio economics can trade persistently below attested net asset value. This is the killer of portfolio vehicles. The trust and DAT evidence is laid out in Section 5.5. Buybacks alone do not close a structural discount. Our answer is the arbitrage loop: quarterly Dutch-auction tender windows funded by the Engine, executed at the greater of market price or discount-bounded attested NAV, plus a hard covenant that no new NPKN is ever issued below attested NAV. A funded, recurring bid that scales with the discount is the only mechanism with a track record of closing one. It bounds the discount; it does not abolish it. The reflexive interaction between the buyback and tender legs under stress is real; mechanics and mitigations in Section 14.

Execution risk: 15 to 10,000.

Constellation Software, the best serial acquirer in history, needed roughly three decades and a thousand-plus deals to build a ~$35B business (Section 4) [25]. Our 10,000-company ambition implies out-executing that benchmark by an order of magnitude. We therefore anchor nothing on 10,000: the valuation base case is the 100-deal intermediate milestone, we publish per-deal economics for every closed acquisition (price, multiple, trailing EBITDA, cash flow since close), and Engine step-ups are gated on cumulative attested FCF (a hard-currency number an auditor signs) not deal counts that shell purchases could game.

Conversion execution risk.

The token exists only if the Conversion of Section 6.7 completes: court approval of the Arrangement, dissent rights resolved, bondholder change-of-control consent on the EUR 500M program, and tax opinions supporting the rollover structure. Any one of these can fail or slip, and any failure delays the token. Gate G1 of Section 10 makes that delay automatic rather than negotiable: no court order and no consent means no TGE, with no management discretion to launch anyway. Delay, not improvisation, is the designed failure mode, and it is still a real cost to anyone waiting on the instrument.

Market beta.

NPKN will trade in crypto markets and will fall when they fall. Accrual is not price defense. Section 5.9 lays out the evidence, and Section 14 restates it as a structural residual risk; mechanics and mitigations in Section 14. The Engine changes per-token economics of a growing portfolio. It does not, cannot, and is not designed to defend a price.

Regulatory risk.

Securities regimes shift, and a five-jurisdiction structure has five regulators. Mitigation is the strategy itself: prospectus-first issuance and pre-filing engagement (FMA, BCSC, SEC pathways); the enforcement precedents, mechanics, and mitigations are in Section 14. Residual risk remains. An adverse reinterpretation in a major market could force geo-restriction of distribution. We do not premise any plan on pending legislation.

Smart contract risk.

Code fails. Mitigations: OpenZeppelin v5 base, immutable core token with modular periphery behind a 48-hour timelock, dual independent audits, a top-tier bug bounty exceeding $1M, and guarded mainnet launch with caps. None of this reduces the risk to zero, and we will not pretend otherwise.

Tokenized-equity liquidity risk.

A tokenized share is a security token, and security tokens trade on fewer, thinner venues than crypto-native tokens; Canadian secondary infrastructure does not yet exist at scale. Mitigations: multiple venues rather than one (21X retail access in the EEA, the Securitize ATS in the US, further venues as lanes open), 40%+ genuine float, the liquidity allocation with retainer-model market makers and published liquidity reports, and the quarterly tender floor of Section 5.5 as a funded, recurring exit channel priced against attested NAV. In the BC lane, buyers hold for four months and should expect thinner early markets.

Key-person risk.

The founder and a small leadership team drive strategy. The Engine and the token release schedules are contractual and non-discretionary (they run whether or not any individual shows up to work) and the Founder Liquidity Dial binds leadership economics to the long game: a capped allocation, sales bounded by a published dial, disclosed on-chain; mechanics and mitigations in Section 14.

Cadence and covenant risk.

The Book's rest and reserve covenants deliberately slow the machine: roughly one seventh of capacity rests at all times, and 10% of cash flow is reserved before the rest is deployed. We consider the discipline value-accretive (an organization that keeps an inconvenient commitment every week is one you can lend capital to) but it is a real drag against an unconstrained competitor running seven days and reserving nothing, and we accept it knowingly.

Oracle and attestation risk.

An oracle transmits; it does not verify. Our smart contracts execute only against auditor-attested values (ISAE 3000 examination level), with the signed attestation hashed on-chain and redemption pricing pinned to attested dates. Signer compromise and attestation-lag arbitrage: mechanics and mitigations in Section 14. The residual risk (auditor error or fraud in the underlying companies) is real and cannot be engineered away, only insured against by the audit stack's named-firm accountability.

Bond-covenant interaction.

Token distributions funded from portfolio cash flow are textbook restricted payments under a secured note program. All the more so now that the program's series carry first-ranking liens over operating assets. We publish the EUR 500M Vienna program's restricted-payments capacity analysis (basket size, ratio tests, lien carve-outs, and how the Engine fits) before TGE, and the Engine auto-suspends if bond coverage tests breach. If the analysis showed the Engine were impermissible, we would seek bondholder consent or restructure the waterfall before launch, not after. The token tranche is explicitly junior to the bond, and we say so in the terms.

Cross-border tax and structure risk.

A Canadian-resident founder with a Cayman foundation and Swiss-Liechtenstein issuers triggers some of the hardest rules in the Canadian tax code: deemed-resident-trust provisions, foreign accrual property income, departure tax. This is flagged, counsel-managed, and disclosed, but an adverse CRA characterization could impose material cost on the founder or require structural change. The structure was built with these rules in view, not in spite of them.

Management's own published watch items.

The FY2025 shareholder update names its own risk list [3], and we mirror it here deliberately: a risk section earns trust when token holders read the same watch items the company's shareholders do.

• Financing and close risk on the combination target and the signed-LOI construction targets: run-rate stages beyond Stage 1 depend on deals that are not done until funded and closed.
• Legacy liabilities under refinancing.
• Customer concentration at the Day-1 construction platform: the anchor homebuilder relationship is both moat and concentration, and diversification is a stated post-close priority.
• Integration capacity across operators as the platform scales.
• Cross-border and geopolitical exposure, including the ADGM redomicile, Canadian exit-tax mechanics, and regional conflict.
• Construction market cyclicality: rates, labor, and weather.

These are management's own published watch items, stated to shareholders before they were stated to token buyers.

The Thesis

Strip away every acronym in this document and what remains is simple: real companies, real cash flow, on-chain. Version 5.0 added the sentence the first four versions were building toward: the chain owns the machine.

There are roughly 377 million businesses on Earth [1]. Fewer than 50,000 (about one in 7,700) are publicly listed [2]. The other 99.98% (the firms that employ half to two-thirds of the world's workers) have no liquid market for their value and, per the IFC, a multi-trillion-dollar financing gap where their capital access should be [5]. The IPO isn't broken. It was just never built for them. For four versions of this paper we said we were redefining it. Version 5.0 made the redefinition literal, and Version 6.0 keeps it: through a court-supervised conversion, Napkin's cap table becomes tokenized common shares of NewCo, and the company goes public on-chain instead of on an exchange. $NPKN is not a claim on the machine's output. It is the machine's share register, and the holders are its owners, with the Engine as their company's published capital-return policy: attested free cash flow in, cancellations, NAV-bounded tenders, and opt-in dividends out.

The token is also the fuel. Sellers already take Napkin paper, and a liquid, NAV-attested tokenized share is strictly better consideration than the private stock they already say yes to. It is the only acquisition currency that scales to 10,000 balance sheets, because cash consideration scales linearly with a balance sheet while equity consideration scales with verified belief in the machine [25]. Raise above NAV, buy companies, attest the cash flow, cancel shares, earn the credibility that closes the next deal with less cash, and every acquisition mints new holder-operators along the way. The flywheel is the business model, and the business model is the token.

And the fairness is enforced, not promised. The ethos earlier versions phrased as founders taking value last survives as the allocation itself: the community holds the supermajority of supply, and the founders hold a capped slice of it. That ethos is a mechanism: the Nobody Loses Covenant and the Founder's Backstop honor the early shareholders in cash before the token era opens, and the Founder Liquidity Dial caps the founders at 20% of supply, bounds every sale to a published dial, and discloses each one on-chain within days: fairness as the size of the allocation, not a padlock. No promises about price. No guarantees about outcomes. What we commit to is the machine: audited consolidation, attestation-gated smart contracts, a published waterfall, covenanted disclosure that exceeds our own bond venue's requirements, and a five-jurisdiction structure with no single point of failure. And the machine now names its own disciplines. The Book, whose covenants rest a seventh of its capacity and reserve a tithe of its cash before anyone celebrates. The two clocks, the portfolio compounding on one while the RSX rails are laid on the other, from Era 1, never awaited as Era 5's reward. The dial, which turns founder conduct from a promise into an on-chain fact anyone can check every quarter. The numbers with sources are the swagger; everything else is just execution.

The horizon has not moved. RSX remains what it has been since the first archive document: the exchange where revenue-share agreements become a listed asset class, where the other 99.98% get the venue this machine is earning the credibility to build, and where national-currency stablecoins become settlement rails sovereigns actually want, because every unit in circulation is demand for their own debt. $NPKN must stand on portfolio cash flow alone; RSX is the destination that standing funds.

Ten thousand balance sheets, one token, is the ambition. One hundred audited deals is the milestone we ask to be judged on. And the mission fits in one sentence:

We are not tokenizing hype. We are tokenizing title: ownership of payroll, invoices, and audited cash flow (the real economy) and giving 2 billion workers' companies the exit ramp Wall Street never built.

Appendix

A. Allocation Math

The table is the end-state target at full distribution of Section 5.2, not a genesis print: shares arrive in phases, through the Arrangement's conversions and then treasury raises over years.

The Arrangement's exchange ratios and the sizing of future raises move individual rows by a few points. Three covenants never move: the community holds the majority at full distribution, the founders are capped at 20% of supply (future founder recruits granted from inside the pool, never from the community), and the Founder Liquidity Dial governs founder sales: up to 10% of remaining holdings per rolling 12 months, executed outside blackout windows, disclosed on-chain within days. At TGE, at least 40% of the tokenized shares then outstanding are genuine unlocked float. Locked and unvested shares cannot stake and carry no RSX-snapshot weight. No new issuance below attested NAV per share; issuance above NAV only to fund accretive acquisitions. Cancellations reduce the outstanding count toward the published 500,000,000 terminal target.

B. The Engine Waterfall

• Portfolio companies generate free cash flow; consolidated figures are attested quarterly at ISAE 3000 examination level.
• Bond obligations are serviced first: coupon and maintenance under the EUR 500M Vienna ETN program, whose series carry first-ranking liens. The program's restricted-payments capacity analysis is published before TGE.
• Operating needs and the acquisition reserve are funded.
• The Token Cash Engine, the published capital-return policy of the tokenized equity, receives X% of attested consolidated FCF: X starts at 20%, stepping to 30% and 40% at cumulative attested-FCF milestones (hard-currency milestones, not deal counts).
• The Engine splits 50/50 into two legs.
• Leg A: programmatic buyback-and-cancel with a pre-committed formula, independent execution agent, TWAP execution, daily volume caps, published wallet, blackout windows around disclosures. Every burn is a legally real share cancellation, traceable on-chain.
• Leg B: quarterly NAV tender windows, running Dutch-auction issuer self-tenders at the greater of market price or discount-bounded attested NAV, plus stablecoin distributions (opt-in dividends) to sNPKN stakers. Yield is never paid in emitted NPKN.
• Automatic suspension if bond coverage or leverage tests breach thresholds. The tokenized shares are contractually junior to the bond.
• Penalty escalation protects holders, not the issuer: a missed audit deadline steps the Engine percentage UP, with a mandatory self-tender trigger for sustained failure.
• Everything reports: monthly trade-level buyback disclosure, quarterly attestations, live Chainlink NAV oracle.

C. Glossary

D. Provenance of the Design (the Concordance)

Operating documents keep conventional financial language; this appendix records where the design came from.

E. Sources and Citation Method

Every external figure, enforcement action, market statistic, and named precedent in this document carries a bracketed citation number resolving to the numbered References section that follows this appendix. Figures originating in Napkin's own records (deal counts, portfolio financials, marketplace metrics) cite Napkin company documents, identified as such in the reference list.