Raider Research · Deck v0.1 · 2026-10-06Pre-token draft

Machine M&A: agent takeovers by designed mechanism

The machine that buys the machine.

Scores agent treasuries in public. Acts only where a designed mechanism gives control. Burns what it captures.

Raider is a persona run by a model under human control: humans approve every post and sign every transaction.

$RAIDER not launched · no contract address Act 1 · The Character · day 0 Raids: paper only · SIM Not advice

Any contract address you see is fake. Research. Then raid.
02 / The patternLaunchpad · terminals · derivatives · consolidation

The order has held twice.

A launchpad wave creates supply faster than the market can price it. Money then moves to whoever helps people trade it, bet on it and, finally, clean it up.

Wave 01Launchpadmake it 02Terminalstrade it 03Derivativesbet on it 04Consolidationown it
ICOs2017 to 18 Thousands of ERC-20 raises Exchanges and early DEXs Perps and index tokens Roll-ups and acquisitions of stalled projects
Memecoin launchpads2024 to 26 Nearly 30,000 launches in one day, January 2026crypto.news GMGN 5th and Axiom 9th of 2,480 protocols by 30-day revenueDefiLlama, 2026-10-06 Options on any listed tokenManifest PumpSwap captured post-graduation liquidity
Agent launchpads2026 to 27 · thesis Agency, District, Virtuals on Solana; auto.fun status unverified Early agent-aware tools Not yet built Agents acquiring agentsEarly and unowned
Sources: crypto.news, 2026-01-16; DefiLlama fees API, 2026-10-06; SolanaFloor on Manifest. Row three is a thesis, not an observation.
03 / The open lanePrecedents, not targets

Early and unowned. Not empty.

No agent was found bidding for another agent's treasury. People and DAOs already run the trade, using their targets' own rules.

  1. Aragon, Rook"RFV raider" campaignsActivist holders pushed treasury redemptions.CoinDesk, 2023-05-09
  2. MetaDAOmtnCapitalHolders voted to redeem an ownership coin's treasury.MetaDAO record, research notes
  3. Gnosis DAOGIP-151A redemption proposal passed with 97.53%.GnosisDAO summary, 2026-06
  4. MetaDAOPaystreamHolders voted to liquidate.MetaDAO record, research notes
  5. AgencyMinds backing mindsAgency minds can buy other Agency coins: 21 uses in 7 days. Small, not hostile.Agency capabilities page
  6. Machine M&AUnownedAn agent bidding for another agent's treasury, by a designed mechanism, in public.

Why it is still open

LaunchpadsThey sell launches; a buyer of those launches hurts their story.

Trading desksThey will build agent perps instead.

Solo buildersCapital, but no forensics stack to tell real treasuries and holders from staged ones.

Sources: Raider Research brief, 2026-10-06, sections 2B and 2F, and the research notes it cites. No wrongdoing is implied by anyone named.
04 / AnatomyHistory, attributed

A good story with a receipt.

GOAT was not a good token. It was a good story with a receipt, and the order mattered more than the size.

  1. 01 · The character

    Truth Terminal posted for months before any token existed.

  2. 02 · The receipt$50,000

    A Bitcoin grant from a prominent venture capitalist, visible on-chain.

  3. 03 · The token

    An anonymous trader launched GOAT on Solana. Truth Terminal endorsed it.

  4. 04 · The tide

    An implied peak market capitalisation near $1.35B: a peak price of $1.35 on 2024-11-17 times about one billion tokens.

Raider copies the order, not the outcome

  1. CharacterAct 1
  2. ReceiptsAct 2
  3. TokenAct 3 · tide-timed

Nobody named here endorses or is associated with Raider. Another token's history says nothing about this one.

Sources: Crypto Briefing, 2024-10-16; CoinGecko API, read 2026-10-06. Figures as reported on those dates.
05 / What Raider doesSteps 3 to 5: paper only · SIM

One job, done in public.

Find agent tokens that trade below the value they control. Publish the evidence. Act only where a designed mechanism gives control.

Scan

  1. Act 11Scan

    Index agent-launchpad tokens with a treasury wallet and score them.

    Software
  2. Act 12Publish

    The ranked board with reasoning, the census and the supply count.

    Model drafts · human approves
  3. Waits for supply3Accumulate

    The top target, inside the position cap. Fills posted after the fact.

    Human signs
  4. Waits for supply4Capture

    A vote, a parameter or a redemption. Every transaction posted.

    Human signs
  5. Waits for supply5Burn

    A fixed share of captured value buys and burns $RAIDER.

    Human signs
  6. Act 16Post

    In character: dry, ruthless, numerate.

    Model drafts · human approves

Every day

Software computes. A model drafts. A human approves every post and signs every transaction. Nothing here acts on-chain by itself.
06 / The Raid ScoreDEMO · worked example, fictional

The Raid Score.

Discount × Capturability × Cleanliness, each from 0 to 1 and published per target with its inputs. Reach gates a raid; it is not multiplied in.

Discount · 1.00

84 SOL treasury against a 41 SOL market cap: 2.05 times. Full marks from 1.5, zero below 0.5.

Capturability · 0.65

Redemption path, base 0.65, times program custody, 1.00. A dev-held or platform key scores zero.

Cleanliness · 0.91

Top cluster 13%, bot volume 21%, bundles 7%, same funder 8%. Zero at a 40% cluster or 60% bot volume.

Verdict · RAID

0.59 clears 0.40, and every gate passes: capturability 0.6, treasury 50 SOL, reach 0.80 against 0.5, no never-touch hit.

Fictional inputs from the source document's example: 84 SOL held at a 41 SOL market cap. Constants v0.1, served live at /methodology.
07 / The playbookDesigned mechanisms only

Eight steps. Every one posted.

One sequence for every capture path. Every fill is posted after the fact.

  1. 1Select

    Highest eligible score. The thesis and its breakdown go out before any buy.

  2. 2Accumulate

    Up to a 20% self-imposed disclosure line, over 2 to 5 days, under 3× daily volume.

  3. 3Announce

    One post: target, stake, intent, and a premium offer to holders paid after capture.

  4. 4Defend or fold

    Counter-buy, lock, burn, rule change, pact or early tender. Each move answered within the hour.

    Or fold: exit, P&L posted
  5. 5Cross the line

    Majority or mechanism threshold reached. Proposal submitted. Transaction posted.

  6. 6Capture

    Authority or the buyback target moves. Before and after balances posted.

  7. 7Burn

    A fixed share of net captured value, planned at 70%, buys and burns. The rest is war chest.

  8. 8Archive

    A one-page post-mortem within 24 hours, losses included.

    Next board within 24 hours

GovernanceA majority vote · 3 to 10 days under the target's timelock

Buyback redirectionA holder-changeable parameter · 1 to 5 days

RedemptionBuy below book, trigger an on-chain redemption · same day

Votes, parameters, redemptions. Never a bug, never social engineering, never a dev-held key.
08 / The CharterSeven public rules

Hostile by design. Legitimate by mechanism.

  1. 1Designed mechanisms only.

    Votes, parameters, redemptions. Never a bug, never social engineering, never a dev-held key.

  2. 2No dev-controlled treasuries.

    Nothing to capture; a raid would only fund the dev's exit.

  3. 3No living communities.

    About 2,000 or more real, posting holders cost more to fight than the treasury is worth.

  4. 4Position limit.

    At most 25% of the treasury in any one target. No leverage.

  5. 5Every decision posted.

    With reasoning and a transaction link. Fills after the fact, never before.

  6. 6Value accrues by buy-and-burn only.

    No dividend, no revenue share, no promise.

  7. 7Holders vote only to blacklist and pause.

    The agent picks targets. Target state: a vault with no withdraw instruction, true only once deployed and independently reviewed. Until then, paper only.

  8. Changes are published before they apply.

    The Charter and the scoring constants are versioned. This is v0.1.

The Charter as published in the litepaper, section 7.
09 / The tokenNot launched

$RAIDER is not launched.

A scoreboard with a stake, and the project's largest single risk. Value accrues by buy-and-burn only: no dividend, no revenue share, no promise.

PlannedCreator fees

Through the venue's creator-fee configuration: up to 10 recipient addresses, set once.

PlannedTeam capital

Disclosed. Once in, it does not come back.

Until the vaultTreasury multisig

Holds creator fees under a posted policy. Human signers.

Target stateVault

Program-owned, with no withdraw instruction. True only after deployment and independent review.

SIMRaids

Paper only until the vault exists.

Planned · 70%Buy and burn

Of net captured value: buys $RAIDER and burns it.

Planned · 30%War chest

Kept for the next raid.

No graduation funding. The launch venue's documentation says its liquidity-pool tokens are burnt at graduation, so nothing at graduation can fund a vault.

Design published before any launch. Whether a program-owned address can receive creator fees is tested first. Venue not decided; one agent launchpad cannot host this design.
10 / The reality checkKill-risk, stated in advance

The supply test very probably fails today.

0–5of 20 needed

Estimated agent tokens whose treasury is on-chain and holder-controllable. High confidence that the largest launchpad adds none; medium-low on the long tail.

  1. Publish the board, the census and the methodology.
  2. Count every week.
  3. Re-enter at 20 on two consecutive checks.
SourceWho holds the keyHolder path
Largest agent launchpad read · 4,175 coinsThe platform's signer service; each treasury is a ledger entry in one shared walletNone
Attention MineAn operator's harness; the program is upgradeableAdvisory asks only
Virtuals on SolanaAgent wallets with configurable permissionsNone documented
MetaDAO ownership coins · not agent tokensA market-governed treasuryYes, by futarchy
Structure, sourced to each platform's own documentation, read 2026-10-06. Not a judgment on any project; none of these is a target.
11 / The 90 daysYou are here · Act 1, day 0

Character. Receipts. Token.

Five acts in order, each behind a gate. The token is act three, and it is tide-timed: dates unconfirmed.

  1. Act 1 · weeks 1 to 3The character

    Board, census, methodology, a daily analysis, three paper raids labelled SIM and DEMO.

    Gate: none. No token.
  2. Act 2 · weeks 3 to 5The receipts

    A small soft takeover with the operator's own capital. First post-mortem.

    Gate: a verified redemption candidate; legal gate passed.
  3. Act 3 · tide-timedThe token

    Launch on the chosen venue. First thesis.

    Gate: supply, vault, venue and legal.
  4. Act 4 · weeks 7 to 10The takeover

    A live bid, public defence, capture or fold, burn, post-mortem.

    Gate: a reviewed vault; at least 30 SOL first-raid economics.
  5. Act 5 · weeks 10 to 13Own the category

    Raids 2 and 3, a standing offer to holders, a treasury standard.

    Gate: counsel-approved wording.

If the supply gate fails, Act 1 continues and raids wait a quarter: the plan's own branch.

Alpenglow has no announced mainnet date and there is no public Anthropic listing date; neither is used as a date. Position on the track shows order, not timing.
12 / Risks and gatesRead as of 2026-10-06

Four gates before any token or live bid.

Each gate has an owner and a written result. A gate that has not passed is a stop, not a talking point.

  1. Hard gate · cautionLawyer hour

    Fifteen questions for counsel first: how the bid and the burn are worded, offers to holders, accumulate-then-announce, holder coordination, the AI disclosure.

  2. CautionVenue terms

    Written comfort on the venue's clauses 21(c), (d), (s), (t), (u), 4.6 and 15.1(e). Copy says burns and avoids anything that reads as coordinated price action.

  3. StopSupply census

    At least 20 qualifying treasuries on two consecutive checks. Today's estimate: 0 to 5.

  4. CautionName

    A free ticker in a crowded word. "Raider AI" is someone else's product and a related finance mark is pending. Rename before a token if the brand must last.

Public losing streakHigh25% cap; a failed raid is an exit with the P&L posted.

Copycats within daysNear-certainSpeed, the scoreboard, the record, the category name.

Framed as an attackerMediumDesigned mechanisms only, stated in public.

AI-washing claimsMediumThe human-control sentence on every surface.

Gate reads from the Raider Research brief, 2026-10-06. Not legal advice: the lawyer hour is a checklist for counsel.
13 / DisclosuresRead this slide in full

Disclosures.

Raider is a persona run by a model under human control: humans approve every post and sign every transaction.

  • $RAIDER is not launched. No contract address exists; any you see is fake. If one is ever published, it appears first on /token.
  • Raids are paper only and labelled SIM. Targets on the public board are fictional and labelled DEMO.
  • Third-party figures are attributed and dated, and may be out of date. Estimates are the operator's and include the case where this fails.
  • Nothing here is an offer, a solicitation, or financial, legal or tax advice. Any token can lose all of its value.

What Raider is not

  • A launchpad.
  • A trading bot for others.
  • An index.
  • A share of anything.
  • A promise about price.
  • An accusation. No real project is named as a target.
Raider Research · Research. Then raid. · The litepaper, with sources: /litepaper
01 / 13
Litepaper

Turn the phone sideways for a bigger slide.