Solana, and how we actually run the launch stack

Short answer

1. Solana runners: Solana made at least 16 tokens that hit $20M within 7 days since Jul 1 (the census had 11). pump.fun made 10, mostly plain memes, and 5 of those are now near zero. StonkFun made 6, all paying holders in a stock or ZEC, and 3 of those still hold $18M to $58M. Momentum is moving back to Solana: pump.fun fees are up 71% in September while Pons (Robinhood) is down 76% and Flap (BNB) down 52% (A).
2. Is Solana a disadvantage? For payout asset, distributor and receipt page: no, it has the best payout-asset menu of any chain. It has two exact disadvantages. Binance Alpha has listed zero Solana tokens since Jan 28, so the second leg must come from the MM partner and tier-2 exchanges (A). And no Solana venue can tax sells more than buys (A, program source).
3. Platform: our own Token-2022 token with a transfer fee, on a Meteora DAMM v2 pool seeded with the MM partner, run by our keeper. It is the only option where the tax binds on every pool, we hold the fee switch (so we can step it down and then provably turn it off), and we can pay in any asset. StonkFun keeps the tax authority and pays the creator nothing in reward mode. Fallback: a pump.fun Custom Pair with its creator fee routed 100% to our distributor.
4. Engine: build once in about 4 to 6 weeks for one engineer (estimate): mint and pool scripts, keeper, Keep-or-Forfeit ledger, pusher, receipt page, monitor. After that each launch is one config file plus a checklist, and about 2 days of our work.
5. Payout assets left: 27 tokenized stocks and other assets are liquid enough on Solana (a $50K swap moves the price 2% or less) and still unclaimed by any reward coin; 70 at a looser $5K bar. About 10 to 15 have a story strong enough to lead a launch. Fame is not required: first plus in the news that week is what ran (DraftKings $5.9M, SanDisk $9.2M; famous S&P 500 and Microsoft firsts stayed under $0.3M).
6. Money: yes, the token buys the stock itself: the tax is sold for the payout asset and pushed to holders. We buy nothing. Our real costs per launch are a few hundred to a few thousand dollars plus an optional pre-funded pot. At 3% tax, a payout big enough to screenshot ($20 a day per $1K held) needs 67% daily turnover, which only happens on days 1 to 3.
7. Pre-funded pot: worth $10K to $25K on a launch whose organic day-1 tax is small. It lifts a quiet launch's day-1 payout by about half (from $45 to $66 per $1K held, estimate) and is noise for a runner.
8. Recommended path: Solana, own Token-2022 mint on DAMM v2. Tax 5% on day 1, stepping down to 0 by about day 14 in steps that snap to Solana epochs. Keep-or-Forfeit carries the sell penalty. Pick a first-in-category payout asset tied to that week's news. The MM partner plus tier-2 exchanges carry days 7 to 14. Launch on BNB/Flap only if the client's plan needs a Binance listing.

Q1. What is the Solana ecosystem doing?

1.1 The Solana runners, Jul 1 to Sep 29

The census found runners through CoinGecko listings, so tokens CoinGecko never listed were missed. We added StonkFun's own pool API (149,612 pools) and pump.fun's API (top 1,050 coins by market cap), then checked every candidate against hourly candles and volume (A, Q1-SOLANA-RUNNERS.md §1).

Rejected after checking: XTAL and VISE (fake prints on almost no volume), CIGR (opened at $15.6M, pre-bought), GP, BTC, PURR, RAYCAT, NEARKAT (too slow or under $20M) (A).

What the Solana set says: - Two families, very different survival. Plain pump.fun memes spike fast (median about 21 h to $20M) and die: 5 of 10 are at about zero. StonkFun reward coins grind (median about 118 h) and hold better: 3 of 6 still at $18M to $58M (A). - First in the payout asset wins, again. ZCAT (first ZEC) reached $150M; SI (first NVDAx) is at its high today; MASK (second ZEC) peaked near $26M (A). - Venues that produced nothing: bonk.fun, Bags, Believe, Heaven, Moonshot, Jupiter Studio, Meteora DBC. Together they take under $1M a month in fees except bonk.fun (A, DefiLlama).

1.2 Odds of a $20M run, by venue

Raw pump.fun launches are mostly bot spam, so 1 in 350,000 is not the odds of a funded, prepared launch. The fair comparison for an agency launch is a curated reward coin: StonkFun 1 in 25,000 raw, Flap tax tokens about 1 in 4,500. Both denominators include thousands of zero-effort launches; a prepared launch sits well above either base rate, but we cannot measure by how much.

1.3 Where the activity is (weekly fees, DefiLlama, A)

The live Solana theme is coins that pay holders in stocks, ZEC or gold (StonkFun: $47.5M paid lifetime across its top 100 coins, B, Stonk Board). pump.fun answered with Custom Pairs on Sep 9 (93 quote assets including stocks) and has a holder-rewards mode that pays holders the creator fee (A, pump.fun docs). No Custom Pairs runner yet.

1.4 Verdict: is Solana a disadvantage for the stack?

Verdict: Solana is the right default. It has the payout-asset menu, the live reward audience and the momentum. Two things change. The MM partner and tier-2 listings are the whole second leg, so price that in. And the sell penalty moves from the tax rate into Keep-or-Forfeit. If a client's plan depends on a Binance listing, launch on BNB through Flap instead.

Q2. Can the five-layer stack run on Solana, and on which platform?

2.1 Venue by venue

Y = native, P = partial or workaround, N = no. All from program source, on-chain reads or primary docs unless marked (Q2-VENUES.md §1).

2.2 Hard limits (the ones that shape the design)

• Token-2022 transfer fee cannot tell a buy from a sell. One rate, applied to every transfer, no exemption list (A, solana-program/token-2022 source). Max 100%; the per-transfer cap maximum_fee can be set to effectively unlimited (A).
• Rate changes land two epochs later. SetTransferFee writes the new rate for current_epoch + 2. An epoch is 432,000 slots, about 2 days, so a change lands 2 to 4 days after the call (A source, epoch length estimate). Only one change can be pending.
• Transfer hooks cannot charge a fee (they get the accounts read-only and can only block a transfer), and DAMM v2, Raydium CPMM/CLMM and LaunchLab reject mints with a live hook; Orca needs a badge (A, source). So there is no on-chain sell tax and no on-chain forfeit on Solana. Forfeit runs in our ledger, with a daily Merkle root posted on-chain so anyone can audit it.
• No Solana AMM charges more on sells. The one direction-aware fee (the DAMM v2/DBC "rate limiter") raises buy fees and is deprecated (A). A pool fee also only binds its own pool; someone opens a cheaper pool and routes around it (the Flap-on-BNB leak).
• Routing works. Jupiter routes transfer-fee tokens both ways and tags ZCAT verified (A, live quote). Phantom holds and shows Token-2022 coins (B). Whether Axiom, Photon, BullX or Phantom show the tax rate is unverified, so the receipt page and caller copy must state it.
• Which launchpads can mint a taxed token: only Raydium LaunchLab. pump.fun create_v2 and Meteora DBC create tokens without a transfer fee (A). This is why StonkFun sits on LaunchLab.
• Payout asset quirks. xStocks are Token-2022 with permanentDelegate, pausable and scaledUiAmount. The issuer can pause transfers. defaultAccountState is "initialized" (A), so new holders can receive pushes, but the pause switch is a third-party risk we must disclose, the same way PAID depended on X Money.

2.3 The single best path

Own Token-2022 mint (transfer fee only: no hook, no permanent delegate, mint and freeze authority revoked) on a Meteora DAMM v2 pool, liquidity from the MM partner, LP locked, with our keeper and Keep-or-Forfeit distributor.

Why, on the evidence: 1. The tax cannot be dodged: it is on the token, so it binds on every pool and every aggregator route (A, ZCAT on 30 pools). 2. We hold the fee switch, so we can run the step-down and then remove the switch forever, which anyone can check on-chain (A). 3. The payout asset does not have to be the pool's quote asset, so we can pick any first-in-category asset, including one listed that morning. 4. DAMM v2 accepts this exact extension set without asking anyone (A), and has a built-in anti-sniper fee schedule (Q3). 5. Direct DAMM v2 launches do get found: NEARKAT launched straight on DAMM v2 and closed at $14.3M (A, census).

What we give up: pump.fun's launch feed and bonding-curve discovery, and exact hour-level tax steps.

Fallback: a pump.fun Custom Pair coin with 100% of its creator fee shared to our distributor wallet. Biggest audience and a trustless fee route. The fee is the same both ways, has no end date, only pump.fun's own pool pays, and the payout asset is limited to pump.fun's list. Later upgrade: our own LaunchLab platform config, which adds a bonding curve to the same taxed token, once Raydium provisions it.

Q3. How we run it, and how it becomes an engine

3.1 The money flow on the chosen path

1. A trader buys or sells. Token-2022 withholds the tax (for example 5%) in the receiving account: the buyer's wallet on a buy, the pool vault on a sell (A, docs).
2. The keeper harvests the withheld tax into the mint (harvest_withheld_tokens_to_mint, anyone can call it) and withdraws it to the treasury (withdraw_withheld_tokens_from_mint, needs our withdraw authority) (A).
3. The keeper sells the harvested tokens through Jupiter for the payout asset (for example SK hynix via SOL). This is the "sell program" every tax token has; selling our own token pays our own fee once more, a leak of about 5% of the tax (estimate). Big swaps run during US market hours: after the close some xStock routes return "Oracle price out of date" (A, Jupiter quote).
4. The payout asset lands in the distributor vault. The ledger splits it (proposal, estimate):
5. half paid out at the next push, hourly, no claim. This keeps the day-1 "it landed in my wallet" screenshot, which a pure 72-hour vest would lose;
6. half accrued to each wallet pro rata to balance, vesting after 72 hours. Any outbound transfer or sell forfeits the unvested part to wallets that stayed. Wallets that bought in the launch blocks or through bundles accrue into the pot, not for themselves.
7. The pusher sends batched TransferChecked transactions to every holder over the minimum ($20 of token, as ZCAT and pump.fun use). It creates payout-asset accounts as needed.
8. Every step is published: harvest, swap, vault balance, each push batch, and the daily Merkle root of the ledger (memo transaction). The receipt page reads all of it.

3.2 Build once vs per launch

3.3 Keys and custody

3.4 Supply, team and MM allocation at launch (proposal, estimate)

• Pool first, no dev bag at launch. Put most of the supply in a single-sided DAMM v2 position starting at a low price (for example $50K to $100K market cap). Meteora supports single-sided launch pools (B, Meteora post), so the launch needs little or no SOL from us.
• MM allocation: a token loan of 1% to 2% of supply with call options is the market norm (B, Flowdesk/Caladan). It sits in a published wallet, excluded from Keep-or-Forfeit accruals.
• Team allocation: at most 5%, locked with a published schedule (estimate). The data says pre-built audiences and big team bags did not help (PATTERNS §1).
• No bundle, no hidden early wallets. Snipers and bundlers collected the most reward in every 2026 case (MECHANISMS A.3).

3.5 Snipers

1. DAMM v2 anti-sniper fee: the pool fee starts very high (up to 99%) and decays on a time schedule over the first minutes (A, DAMM v2 source). That fee goes to the LP position; we route it publicly into the pot, so snipers fund the holders.
2. Launch-block and bundle flagging in the ledger: wallets that bought in the first blocks or in the same bundle accrue into the pot, not for themselves.
3. Keep-or-Forfeit: a sniper who dumps forfeits every unvested accrual.
4. Secret launch time: the pool's activation point is not announced until go.

3.6 Tax schedule on Solana (epoch-snapped, proposal)

Why it runs this long: the Solana reward runners crossed $20M at a median of about 118 hours, and ZCAT did most of its taxed volume on days 6 to 13 (A, Q5). Stepping the tax down by day 7 would have cut ZCAT's 14-day holder payout from about $4.1M to about $3.1M (estimate, q5_model.py).

3.7 Cost of pushing payouts on Solana (A inputs, measured by simulation; SOL $119)

Rent for new accounts is 99% of the cost, 0.00156 SOL per new holder of an xStock (A). The mechanism page's "0.002 SOL per holder" is out of date. Hourly pushes to 20,000 holders cost about $28 a day (estimate). ZK-compressed airdrops are much cheaper but likely do not support xStocks' scaled-amount extension, and they weaken the "landed in Phantom" screenshot (B), so they are not for v1.

3.8 Launch day runbook (one launch)

1. T-7 days: pick the payout asset from the refreshed inventory (liquid, unclaimed, in the news). Book the MM partner's day 7 to 14 plan and the tier-2 listing applications. Client signs launch.json.
2. T-3 days: dry run on devnet with the same config: mint, pool, keeper, 50-wallet push, receipt page on a staging URL.
3. T-1 day: create the mainnet mint (fee config and withdraw authority set, metadata on the mint), mint the supply, revoke mint authority. Set up the Squads multisig. Create the DAMM v2 pool with a future activation point and the anti-sniper schedule, and lock the LP. Publish the receipt page with authorities, schedule and allocations.
4. T0: pool activates. Keeper on. First push within the hour. Queue SetTransferFee(300).
5. T+1h to T+72h: hourly pushes. Milestone cards ($10K, $100K paid). Watch the monitor.
6. About day 3 to 14: fee steps as scheduled; each step posted as its own moment. MM second leg days 7 to 14.
7. About day 14: rate 0, fee authority removed on-chain, final vesting pushes, and a "tax is over" receipt card. The token is now a normal token with a clean record.

3.9 What breaks

3.10 What makes launch #2 to #N cheap

• One launch.json per launch; the scripts read it. Nothing is hand-edited.
• The payout-asset inventory refreshes daily (quote depth, claims from the Stonk Board and our BNB scan), so picking an asset takes minutes.
• The receipt page is one template with a skin.
• A one-page checklist (3.8) with a sign-off line per step.
• Devnet dry run from the same config.
• Per-launch labour after the engine exists: about 2 days of our time, plus monitoring (estimate).

Q4. Payout assets nobody pays in yet ("first-time stock")

Full list: practical/data/payout-assets.csv (200 rows: asset, issuer, Solana mint, pool liquidity, claimed or not, fame note, source). Analysis: Q4-PAYOUT-ASSETS.md.

4.1 What exists on Solana

Supply is huge and liquidity is thin: about 1,080 of the 1,127 xStocks have under $1K of pool liquidity (A, DexScreener).

4.2 Runway

Taken on Solana already (53 payout assets across StonkFun, B): ZEC, NVDA, gold, SpaceX, S&P 500, Microsoft, BTC, ETH, DOGE, HYPE, NEAR, TAO, XMR, SanDisk, DraftKings, Kalshi, Anthropic pre-IPO and about 35 memes and protocol tokens.

Open on Solana (liquid at $50K): QQQ, Micron, TRON, Circle, Intel, Tesla, SK hynix, Strategy, Eli Lilly, Robinhood, Marvell, Nebius, Trump Media, Meta, Take-Two, Alphabet, GameStop, McDonald's, Apple, Amazon, Coinbase, Hims & Hers, Palantir, Berkshire, Mantle, Ethena, the memory ETF (DRAM). QQQ, Apple, Alphabet, Meta, Tesla, GameStop, SK hynix and Trump Media were used on BNB, which weakens "first" with callers who watch both chains.

Runway: about 27 launches that are tradeable, of which about 10 to 15 have a story good enough to lead a launch this month (estimate). The best now: the memory trade (Micron, SK hynix, DRAM ETF), Robinhood (its chain's gas subsidy ended on 09-29, B), Circle and Coinbase, Palantir, Tesla and Meta on Solana, GameStop, XRP and TRON (other chains' retail bases).

4.3 Is the idea exhausted? Does it need a famous stock?

• Not exhausted, but thinning. Each new "first" is smaller: since mid-September SI (NVDA, $45M), then DraftKings at $6M, and $0.3M to $2M for the rest (A/B). StonkFun's daily fees halved in a week (B).
• Fame is neither needed nor enough.
• Lesser names that were first and in the news ran to $5M to $9M: DraftKings (ALLINU $5.9M), NEAR (NEARKAT $5.6M), SanDisk (SNDK $9.2M, during the memory-chip run) (A).
• Famous names that were first with no news stayed tiny: S&P 500 $0.28M, Microsoft $0.27M, ETH $0.18M, TRUMP $0.16M, BTC $1.54M, SpaceX on Solana $1.69M (A).
• The same asset can fail on one chain and win on another. NVIDIA flopped four times on BNB and in nine pump.fun pools with no payout, then ran to $45M as SI, the first NVDAx reward coin (A/B).
• The rule: first payer on that venue, plus an asset in the news that week, plus a venue whose wave is still rising (Q4 §c).

4.4 New first-in-category classes we can create

Hard constraint across all of these: at 3% tax on $1M a day of volume, the asset must absorb about $30K of buys a day (estimate). That limits the menu to the names in 4.2 plus whatever lists next.

Q5. Money: what it costs and who pays

5.1 Is the token buying the stock itself? Do we buy anything?

Yes, the token buys the stock. Every trade pays the tax in our token. The keeper sells that token into the pool and buys the payout asset with the proceeds, then pushes it to holders (flow in 3.1). Traders fund everything.

We buy nothing by default. Our out-of-pocket items are: - the SOL for creating holders' payout-asset accounts (which can also be paid from the tax); - small launch costs; - an optional pre-funded pot.

Liquidity is the MM partner's.

5.2 What a holder sees, launch day vs day 7 to 14 (grade-A volumes, 3% tax, estimate)

The model matches reported payouts within about 3% to 14% (MARSCOIN $258K computed vs $245,943 claimed at T+29.5 h; ZCAT $2.89M vs CoinDesk's $2.8M by 09-07) (A/B, Q5 §1.4).

How big must a payout be to screenshot? (estimate) - Per wallet: about $20 a day per $1K held. At 3% tax that needs 67% of the market cap to trade every day. Only days 1 to 3 do that (ZCAT 259%, 135%, 57% turnover; MARSCOIN 182%, 148%, 46%). ZCAT's average payout was only about $6 (470,000 payouts, B), so frequent small drops landing in the wallet also worked. The screenshot is the asset arriving, not the size. - Pot counter: about $100K paid in the first 24 hours. Both runners were at $82K to $103K by then, and MARSCOIN's viral post was the $246K counter at T+29.5 h.

5.3 Is seeding the pot ourselves worth it?

A $25K pre-funded pot paid over the first 72 hours adds about $8.3K a day.

Verdict: seed only launches expected to be quiet, where it lifts the payout past the screenshot bar and pushes the day-1 counter toward $100K. For a runner it is noise. Budget $10K to $25K and pay it in the payout asset itself, so the first push lands in the first hour (estimate).

5.4 Per launch: small, medium, large

Tax schedule from 3.6 (5%, 3%, 1.5%, 0), 10% of the tax to the agency, 90% to holders. Scenarios: small is our assumption for a launch that does not run; medium replays GP's actual first 14 days (peak about $12M); large replays ZCAT's actual first 14 days (q5_model.py, estimate on grade-A volumes).

What the agency earns (proposal): - a launch fee of $25K to $60K, in line with lean launch packages (C, coinlaunch.space); - plus 10% of the tax while it runs.

The 10% is 0.3% to 0.5% of volume. Flap lets launchers take up to 6% of the tax as commission (A-docs), so this is in range. On ZCAT-size volume the share would be about $455K over 14 days; on a quiet launch it is noise, which is why the flat fee matters. Everything flows through the published treasury, so the agency cut is visible on the receipt page. A hidden cut is what cost StonkFun trust ($1.41M of reward money to its own wallets, B).

The engine build (about 4 to 6 weeks, estimate) is a one-time cost of about one engineer-month, paid back by the first medium launch.

Open unknowns, and how to close them fast

Sources

Per-question source lists with URLs and file names are at the end of each working file: - Q1: StonkFun pool API, the Stonk Board, pump.fun API, CoinGecko Pro on-chain, Binance Alpha token list, DefiLlama. - Q2: solana-program/token-2022 and transfer-hook source; MeteoraAg damm-v2 and dynamic-bonding-curve source; raydium-cp-swap, Raydium SDK, Orca Whirlpools source; pump.fun public docs; Helius RPC reads of ZCAT, KNOTS, ALLINU and NVDAx mints. - Q4: Backed API, Jupiter token and quote APIs, DexScreener, the Stonk Board, our BNB Flap scan. - Q5: CoinGecko Pro OHLCV, Helius rent, priority-fee and simulation reads, Jupiter quotes, Solana fee docs, Cloudflare Pages limits, agency and MM pricing pages.

xAI was not used (region-blocked). All work was read-only: no transactions, no messages, no posting.