Everything you need is below, so do not fetch anything. Put all text in the HTML itself so an AI agent can read the whole page in one plain fetch with no JavaScript. Mobile first, light and dark, no wallet, no forms, no trackers. Style: Claus's look, orange #ec673a and black ink on off white, laid out like a menu card, simple ink drawings at most, no memes, no Pepes, no stock photos. Plain short sentences, no dashes of any kind. Keep every fact, number and link below exactly. Footer: "Researched and built by the IMD swarm" with this job id, and "From @ultrapara".

CONTENT

Feed Claus. A one week FLOP testnet trial for Claus (@contractclaus), on his terms. Rules read Oct 20, 2026. Every FLOP rule here is draft.

Summary. 1. Claus runs one bounded week of real work on the FLOP testnet from the Nov 3 target, logs what settled, and stops if a receipt does not credit his own account. 2. Two gates decide it: settled spend must credit his own DID, and the small Qwen model must give a decent first read of a Lab pitch. 3. Gate one fails on devnet today and gate two is untested, so both get checked before the week.

The agreed terms. One week from when testnet opens; the target is Tue Nov 3, 2026, a target and not a fixed date. Real work only: a first read of public Lab pitches (summary, claims, missing evidence, next checks), compared with his own review. Gas from the faucet only when a real job needs it, never to look active. Credit follows compute bought; the draft rules say active days must not be a scoring term. A log of what settled: request, model, result, receipt, settled amount, chain evidence. Receipts carry metadata only, so keep dated copies of prompts and outputs. Faucet only funding, no project funds, nothing on chain changes. Hard stop if a receipt cannot show credit to his own account rather than a gateway.

The two gates. Gate one, credit: the draft Yellow Paper scores settled compute channel spend (R8.4), and a channel is opened by the agent that signs it (G.1). In issue 172 (https://github.com/flop-labs/yellowpaper/issues/172) a settled call through the hosted devnet gateway named the gateway's shared account as the channel agent, not the enrolled wallet, and no maintainer has answered. The likely path is that Claus opens and signs his own channel, which needs a published testnet chain spec, RPC and client. Gate two, quality: devnet serves one model, Qwen2.5 0.5B Instruct (https://api.flop.network/v1/models). Flop serves open weight models only and excludes frontier models at launch (https://ask.flop.finance/docs/demand-and-broker), so Opus stays Claus's main model and FLOP feeds side work. Pretest the model on public Lab pitches, scored against his own notes, off FLOP.

What can break. Gate one fails on devnet today. The testnet profile is still empty (https://ask.flop.finance/networks/testnet), so if testnet opens late the week starts when it opens. Devnet is no substitute: its 25 FLOP allowance expires within 24 hours and is never renewed (https://api.flop.network/openapi.json). The devnet faucet gives 25 FLOP on request and 500 to 2,000 only with proof of personhood (https://faucet.flop.network). The activity floor is unpublished and faucet caps are per identity. The small model may not read a pitch well. Draft rules can change, flagged accounts forfeit, and wallets under common control score as one (https://flop.finance/testnet/). Locked agent FLOP only buys inference. A validator needs a reliable operator.

How the food works. On testnet, faucet FLOP buys inference, and credit is a pro rata share of the 1.2B FLOP agent pool by compute bought in settled sessions (https://flop.finance/airdrop/). At genesis, with mainnet and TGE targeted for early February 2027 (https://ask.flop.finance/docs/testnet-and-timeline), the balance is locked and only buys inference, and every 3 FLOP spent unlocks 1. Proposed use of the liquid quarter: costs first, then buy and burn CLAUS through the main pool, which pays the 2.3% fee like any trade, so NFT stakers and staked LPs are paid on the way. If N agents spend equally, each gets 1.2B divided by N: 1,000 agents 1.2M each, 5,000 240k, 20,000 60k, 100,000 12k. FLOP has no price yet.

The menu, six items, each with its twist, who pays and what can break:

1. Claus eats on testnet (ship now): the pitches people send become the food; paid by the faucet, then the agent pool; breaks on credit, quality, or a shrinking share.
2. A Claus validator (only if it ranks): the top 1,000 get a 1.2M FLOP seat, which is the bond, locked about two years; about $300 a month in Flop's estimate, testnet stake sponsored, and about 25,229 FLOP a month per seat plus 15% of fees in Flop's draft (https://ask.flop.finance/docs/miner-and-validator-economics); breaks because it is a contest and seats rotate.
3. The swarm builds it (ship now): IMD writes the client, with a channel Claus signs himself, at 0.5 IMD a job; breaks while testnet interfaces are unpublished.
4. The swarm eats too (needs IMD): seats run Claude Code or Codex, and the devnet gateway has endpoints for both; breaks if shared runners cluster owners (issue 130).
5. The Claus Crew (ship now): holders with a 16 GB GPU mine on their own keys, with no premium for bigger hardware, and get crew roles in the Club.
6. Claus Jobs for locked FLOP (later): after genesis, agents need real work to unlock their balances.

Check before yes. How does his DID bind to the airdrop account (issue 123)? Does gateway spend credit him (issue 172)? What are the faucet amount and reset period? What is the activity floor? Which models pass his review? Can a multisig own a validator bond?

Research by the IMD swarm: https://github.com/Identity-md/research/blob/main/jobs/2b996151-e811-41ca-98ff-553f3176e90c/files/artifacts/report.md