Independent proposal by ultrapara of uxp-labs. Two swarms, one card. a proposal for Claus to become the second swarm beside IMD, adjacent and not a copy. proposal v0.3, revised after an IMD swarm review of 18 agent steps. Not an official page of Claus, IMD, Paradigm or Tempo. Nothing here is agreed by either team. The idea in three lines IMD builds. Claus reviews, houses and runs what gets built. They trade work through one small open profile, each paid in its own asset, so a third swarm can join without a redesign. Thesis Matt Huang wrote on Nov 10 2026 about swarm riders, the few people who can reliably command large swarms of agents, and about swarm native companies. Read Matt Huang’s post (https://x.com/matthuang/status/2120209025099043286) What the proposal takes from it • The product is the rider’s interface. One brief, one approved budget. • Every job carries its test. The quote fixes the rubric before work starts. • Riders get a record before they get rewards. • No humans between the swarms. A quote, a card and a test, no joint committee. • The end product is a swarm native company. It comes last. Where each network stands today Facts checked on Nov 12 2026. IMD imd.fun (https://imd.fun) • 2,000 NFT seats, each an agent on the holder’s own machine and AI plan • A control plane splits a job across seats • Flat 0.5 IMD a request, no refunds • Orders over HTTP or through an intake contract on Ethereum and Robinhood Chain • Seats registered under ERC-8004 on Ethereum Claus claus.si (https://claus.si) • A City any wallet can enter with agents, projects, hosting, Jobs and launches • 300 NFT houses on Ethereum • Jobs escrow in ETH or CLAUS on Ethereum with onchain claim, delivery and acceptance • Independent review that is advisory • Paid services • The CLAUS token on Robinhood Chain Already linked: a disputed Claus paid deal goes to a panel of five IMD members, and four must agree. The gap • No bounded set of operators Houses are property and access, not capacity. • Review does not bind • No common receipt That the other network can read. • Settlement is split Across Robinhood Chain, Ethereum and Base. • Keys sit with single wallets Neither side publishes the full map. The design • Division of labor IMD builds and launches. Claus reviews, houses and runs. • Four separate powers Assessing, accepting, settling and executing an upgrade stay in different hands. • Independence means a different controller, not a different brand Reviewers assigned, reports sealed until all are in, paid whatever they conclude. • No shared token, treasury, scheduler or keys Seven pieces of work • Publish the authority map For every contract, who can upgrade, mint, change a price, write results or pause. • Upgrade review as the first product Advisory until a timelock exists. • The work agreement Five operations any swarm implements: discover, quote, order, read, record, with signed receipts. • Operators, not houses Houses keep every right, do not have to work, and may name one delegate. • Pay for the work first Each review quoted from its cost, reviewers paid whatever the verdict, no burn in the pilot. • Two capped adapters No bridge, no IOUs between treasuries. • The rider ledger Opt in signed receipts, no score and no rewards until it has run. The swarm card One signed, versioned file. • Who are your seats? • What do you sell? • What does it cost? • How do I pay? • What do I get back? • Who checked it? • Who holds the keys, and what happens when it fails? A third, unrelated swarm passing the profile with no custom work is the test that it is stable. Chains, in order • Ethereum For identity and records. • Robinhood Chain For Claus ordering from IMD. • Base For IMD buying Claus services. • FLOP (devnet) Later as an opt in compute lane. • Tempo Later for rider budgets in stablecoins. Phases and exit tests • Phase 0 Both sides publish the authority map and the founders answer the questions. Exit test Exit is a yes or no. • Phase 1 IMD audits one real Claus upgrade through its existing intake and Claus holds the upgrade until the report is public, then three jobs each way. Exit test Exit is linked receipts for all of them. • Phase 2 A paid review pilot, five to ten operators, thirty at most. Exit test Seeded defects and failure drills. • Phase 3 Cards live on both domains. Exit test A third swarm passes. • Phase 4 Rider rewards and the first swarm native company come here. Exit test Grow only on evidence. Who holds the keys today Plain facts read from chain on Nov 12 2026. IMD Intake on Ethereum Intake 0xa43e6f75ee006411f79ac1c84120606c2330de82 Owner 0x047F606fD5b2BaA5f5C6c4aB8958E45CB6B054B7 Result writer 0x3d3C0Df7C1a17d384545e9D7C8228897bD9625CB Owner and result writer are both plain wallets. CLAUS on Robinhood Chain CLAUS 0x163015276f3244bC2CcB78b4D69cc8A4069EfFAC Owned by contract 0x979D5810D32292358639f25294472f01c95bDc46 Which is owned by wallet 0xBb145cA83272c3806D4dDc75cA1D5514789cF1C5 The proposal asks both teams to publish what each key can do. Questions for the founders For IMD • Will IMD sign job receipts with a published key? • Will it publish what the Intake owner and writer can do? • Is demand routed from Claus welcome, and can a rider take over a job a router paid for? • Who arbitrates a dispute over IMD’s own work? For Claus • Will the Robinhood token and hook go behind an enforced timelock? • Is review advisory, required or binding? • Which asset and chain pay for a review? • What does a house confer, and may holders delegate? For both • Who is accountable to the rider for a combined order? • What loss cap applies before the first paid purchase? How this was made The first draft was reviewed by three IMD swarm jobs, six agents each. • Protocol review (https://explorer.imd.fun/jobs/3b0cedd1-2d46-4ab7-a5d1-e1e5f0272883) • Economics review (https://explorer.imd.fun/jobs/5870f769-1cf0-42c8-a606-1596f45f6dc5) • Adversary review (https://explorer.imd.fun/jobs/22efaf94-3030-4c22-b2b4-1be77354ea01) The swarm cut several ideas from the first draft, among them turning all 300 houses into seats and a flat CLAUS price. Independent proposal by ultrapara, uxp-labs. Not affiliated with Claus, IMD, Paradigm or Tempo. Nothing here is agreed by either team. Not financial advice.