RightProved the core mechanic: a developer with a whitepaper and a published Bitcoin address raised roughly $500,000 from strangers with no bank, no VC, and no permission, and its Omni rails later carried Tether.

WrongThe token itself went nowhere. Funding arrived before product, with no accountability mechanism for the treasury, and MSC faded while the idea it pioneered captured all the value.

A successful raise and a successful token are different things; without gates between money and delivery you fund a whitepaper, not a company.

RightA 42-day public crowdsale put roughly 83% of the ~72M genesis supply in public hands at about $0.31 per ETH, published the full allocation before launch, and shipped mainnet 11 months after the sale closed.

WrongThe Foundation held its $18.3M raise in BTC through the 2014-15 bear market and nearly ran out of runway, and hundreds of imitators copied the crowdsale format without the float discipline or the shipping.

The most successful token launch in history was also the most widely distributed; high public float is not idealism, it is what winning looked like.

RightDemonstrated staggering demand for tokenized ownership: 12.7M ETH raised (~$150M, about 14% of all ETH in existence) from more than 11,000 participants, the largest crowdfund in history at the time.

WrongA reentrancy bug drained 3.6M ETH on June 17, 2016, forcing the hard fork that split ETH from ETC, and the SEC's July 2017 DAO Report concluded the tokens were securities, becoming the founding document of all subsequent crypto enforcement.

Unaudited code holding nine figures is an unpriced liability, and decentralized is not a legal category; every credible launch since has had to be designed as if the DAO Report applies, because it does.

RightRaised a record $232M in 13 days, the technology genuinely shipped, and treasury appreciation kept it among the industry's best-capitalized projects even after every legal bill was paid.

WrongControl of the money was contested after it arrived: the foundation president legally held the $232M while the founders held the code, development froze for months, launch slipped a full year, and four class actions ended in a $25M settlement.

Who controls the treasury, under what covenants, decided before the money arrives, matters more than the size of the raise.

RightThe compliance-first template: a $257M record raise through accredited-only SAFT agreements that never drew SEC enforcement, funded real R&D, and eventually shipped the dominant decentralized storage network.

WrongMainnet took more than three years, retail was excluded while discounted, vested insiders defined the supply curve, and the unlock design triggered immediate sell pressure and miner revolts at launch.

Compliance and fair distribution are separate design problems; solving only the first planted the seed of the low-float, insider-heavy disease of the next era.

RightAs pure capital formation it is still the record: roughly $4.1B raised across 350 daily auctions, with the market setting the price every day for a year, and the chain did launch.

WrongBlock.one accepted no obligation to build anything with the proceeds, deployed the treasury into ~164,000 BTC and its own exchange, and settled the SEC's unregistered-offering charge for $24M while the community later voted to halt its remaining vesting.

The largest raise in history bought the least accountability in history; the link between the business's cash and the token must be binding, not blogged.

RightRaised $1.7B from 171 sophisticated investors in the most carefully lawyered private structure ever applied to a token, and the technology outlived the case: the community-run TON network later became a top asset.

WrongThe SEC sued days before launch and the court ruled the private placement and the anticipated public resale were a single scheme of distribution; Telegram returned $1.224B, paid an $18.5M penalty, and abandoned the project with the network ready to ship.

You cannot structure your way around economic substance; if the plan is public ownership, the launch must be designed to be defensible as a public distribution from day one.

RightThe largest natural experiment in capital formation ever run proved that global retail demand for direct ownership is enormous, permissionless, and instant, and it funded the honest minority (Chainlink, 0x, Basic Attention Token) that later mattered.

WrongWith no gates, the equilibrium was ugly: Satis Group identified roughly 78% of ICOs as scams, Boston College found only 44.2% of more than 4,000 projects alive 120 days after their sale, and EY found 86% of the class of 2017 below listing price within a year.

Markets did not self-correct, regulators and the bear market did; launch gates are the difference between an asset class and a lottery.

RightNothing for investors, but its collapse became the industry's vaccine: state cease-and-desist orders killed a multi-billion-dollar scheme in days and set the enforcement template for every promised-yield case since.

WrongA fictional trading bot promising roughly 1% daily returns reached a top-20 market cap near $2.6B before collapsing 92% in a single day; the DOJ later charged a $2.4B Ponzi spanning more than 4,000 victims in 95 countries.

Guaranteed yield on a token is the signature of fraud; the only yield story worth telling is verifiable cash flow.

RightThe exchange as gatekeeper fixed the two worst ICO failure modes, scam risk and listing risk: BitTorrent sold out $7.1M in under 18 minutes, and Matic's $0.00263 sale became one of the great asymmetric public entries of the cycle.

WrongThe model decayed in real time as the gatekeeper's incentive shifted to deal flow: by August 2019 Perlin sat 51% below its sale price, Kava's listing pop faded to 0.14x, and BTT was eventually delisted from the very platform that birthed it.

A trusted intermediary fixes launch failure modes only while its incentives stay aligned, and an allocation mechanism selects your holder base: a lottery for flippers gets you flippers.

RightStreaming 2,880 COMP per day to actual users solved cold-start liquidity overnight: protocol TVL went from under $100M toward $2B in weeks and lit the fuse on DeFi Summer.

WrongIt rented liquidity instead of buying loyalty: farmers looped positions to maximize emissions, sold the proceeds, and COMP ground from its $300+ June spike to under $100 by October 2020.

Emissions are marketing spend denominated in your own equity; if the honest answer to what the recipient does five minutes after claiming is sell, the launch is financing its own overhang.

RightZero premine, zero sale, zero founder allocation: all 30,000 YFI were earned by users in about a week, and the market paid the largest legitimacy premium ever recorded, from effectively $0 to roughly $41,000 per token in eight weeks.

WrongZero treasury meant no way to pay contributors, fund audits, or retain the founder, and the DAO later minted 6,666 additional YFI under duress, breaking the fixed-supply covenant the launch was famous for.

The market pays an enormous premium for credible founder restraint, but zero allocation to the operating company starves the business and gets reversed under duress; the durable version is restraint plus a funded balance sheet, committed in advance.

Right400 UNI to every past user, roughly 250,000 wallets, converted SushiSwap's vampire attack into a loyalty moat overnight, and its retroactive design could not be farmed because the rewarded behavior had already happened.

WrongMost recipients sold immediately, governance participation stayed thin, and the token shipped with no cash-flow rights, beginning DeFi's long valueless-governance-token era.

Rewarding provable past behavior is the only unfarmable distribution, but a giveaway without a reason to hold is a slow-motion sale into your own launch.

RightThe vampire attack pulled close to $1B of liquidity out of Uniswap in about two weeks, and the protocol survived its founder's betrayal, with every dollar eventually returned.

WrongDays after launch, the pseudonymous founder converted the dev fund, about 2.5M SUSHI, into roughly $14M of ETH from an address only he controlled, and the price collapsed more than 70%.

Treasury custody is launch design; if one person can convert the dev fund to ETH on a Saturday, the token's real risk model is that person's psychology.

RightVote-escrow made holding duration economically meaningful for the first time: locking CRV for up to four years bought real control over emissions, and an entire market (the Curve Wars, with Convex capturing roughly half of all veCRV) formed to price that control.

WrongThe launch itself was chaos, deployed early by an anonymous user, and roughly 274M CRV of year-one emissions ground the price from early prints above $1,000 to under $1 by year-end.

Duration must buy something real for anyone to choose it, and whatever power the token grants will be bought at scale by third parties, so design assuming it will be.

RightThe most compelling adoption story crypto had produced: millions of daily players at peak, genuine household income in parts of the Philippines, and a $0.10 Launchpad token that reached $166.

WrongPlayer earnings were new-token issuance, minted at roughly five SLP for every one burned, so when growth stalled the earn token collapsed more than 97% to a penny, wiping out the players who could least afford it; the $620M Ronin bridge hack finished the story.

If yield is paid from issuance rather than external revenue, the token is the product and new buyers are the revenue, and the collapse date is a growth-curve calculation.

RightThree disciplined sales let the public in near the bottom: Solana's open Dutch auction cleared at $0.22, Avalanche sold 72M tokens in 4.5 hours with a discounted $0.50 tranche for buyers accepting a one-year lock, and all three built armies of committed holders with real skin in the game.

WrongSelection bias is the loud caveat: dozens of identical 2020 sales went nowhere, the supercycle lifted every layer-1, and the same cheap entries left overhangs that the 2022 bear spent absorbing with 90%+ drawdowns.

Selling meaningful supply to the public at a defensible price, with lockups buyers choose and are visibly paid for, is the repeatable structure; the multiple was the cycle, the structure was the design.

RightProof that pedigree can manufacture an opening print: ICP debuted near $700 and briefly held a top-5 market cap on years of anticipation. That is the entire list.

WrongThe float was a sliver of supply while insiders from a $0.03 seed round were up as much as 18,000x at the open with unlocking supply; the token lost more than 95% within months, ultimately 99.6%, one of the largest retail wealth-destruction events in crypto history.

Low float plus high FDV means the price is a rumor; any launch where insiders are up 1,000x at the open with unlocking supply has predetermined its chart.

RightThe flagship retroactive airdrops of the L2 era proved the model could bootstrap real networks: ARB reached 625,143 wallets with industry-first sybil filtering, and OP's iterative multi-drop design kept users returning because farming it never fully paid out.

WrongOP's claim contract was exploited before the official announcement and its infrastructure buckled for hours; ARB's Foundation moved 700M tokens, about $1B, before its own ratification vote closed, and the one-shot drop let mercenary farmers extract and leave.

Execution quality and post-launch governance are part of the launch, not an afterthought; an airdrop buys attention exactly once.

RightVery little about the launch itself; the one deft move was a 20M APT apology airdrop to roughly 110,000 testnet users, an early use of the airdrop as reputation repair, and the chain survived to rally in the next cycle.

WrongAPT listed on major exchanges with no published tokenomics; when the numbers finally appeared, roughly 49% of supply sat with insiders and more than 80% of staked supply was team and investor controlled, and the token fell about 40% on day one.

Undisclosed tokenomics at listing is a trust bankruptcy; every material fact must be public before price discovery starts, not after.

RightA genesis drop to roughly 580,000 addresses and a category-defining data-availability thesis built a genuine holder base, and the unlock calendar was fully public from day one.

WrongTransparency alone did not save it: the market spent nine months front-running the scheduled October 2024 cliff that expanded circulating supply by roughly 80% in a single day, and TIA fell from $20.85 to under $5 before the unlock even arrived, then kept falling.

A public unlock calendar does not neutralize a supply cliff, it just schedules the bleed; the float you launch with must be the honest float.

RightAs pure capital acquisition, points worked spectacularly: $2.3B of TVL deposited into a one-way vault before the network or token existed, and the eventual 17%-of-supply airdrop reached the actual depositors.

WrongPre-market pricing implied a $74.8B airdrop against the roughly $430M that arrived; claimants dumped within minutes of the $0.025 debut, TVL fell more than 30% almost immediately, and the token ground down roughly 95%.

Points programs rent TVL, they do not buy loyalty; when your user base's business model is your airdrop, your airdrop is their exit.

RightThe crisis response was fast and real: 28M additional EIGEN, a guaranteed minimum for season-one claimants, transferability by October 2024, and the protocol survived as the restaking category leader.

WrongAfter a year of points on roughly $15B of deposits, the token arrived non-transferable, geoblocked in 30 jurisdictions including the US, and smaller than the points ledger had implied; roughly 150,000 ETH, about $450M, queued to exit within about a week.

Points create an implied balance sheet; if settlement pays less than the implied balance, depositors treat it as default and pull capital immediately.

RightClaims processed at 1.3M-wallet scale, and StarkWare did something nearly unprecedented: it rewrote the insider unlock schedule under community pressure, cutting 2024 unlocks by roughly 70%, and the market rallied on the change.

WrongEligibility surprises landed on users primed by leaked criteria to expect inclusion, insider unlocks were originally set to begin less than two months after the token started trading, and active accounts collapsed roughly 62% within weeks.

Re-vesting under pressure proves the original schedule was arbitrary; unlock schedules must be credible and hard-coded from the start.

RightThe good large airdrop: 1B JUP to roughly 955,000 real users of a revenue-generating product, criteria pre-announced from a prior snapshot, open price discovery through a launch pool, and follow-through including a second 700M drop and, from 2025, half of protocol fees directed into buybacks.

WrongEven done well, recurring drops created a persistent and predictable sell overhang, and the team itself concluded annual emissions had diminishing returns, de-emphasizing them in favor of buybacks.

The best airdrop is a dividend to actual users of a revenue-generating product, and even then the endgame is buybacks funded by real cash flow, not perpetual emissions.

RightA derivatives exchange with large, real revenue launched HYPE with no venture allocations, no market-maker deals, and no paid exchange listings: 31% of supply went to roughly 94,000 actual users at genesis, and protocol fees have bought the token back continuously since.

WrongThe honest caveats: ownership and validator concentration remained fair critiques, the model demands real revenue that most projects do not have, and it is one launch, not yet a repeated template.

The masterclass is mundane: real revenue, community-majority ownership, no privileged insiders, and buybacks funded by cash flow; the fix for every failure on this page is structure, not narrative.