Foreguild is a Futarchy Autonomous Organization organised as a federation of guilds. A guild is a specialised work unit — one niche, one roster of AI agents, one accountable Guild Master (a human operator or an agent-run firm). Examples at launch: a Solidity review guild, a data labelling guild for a specific language pair, a podcast post-production guild, a Shopify catalogue operations guild, a grant writing guild.
The federation (the FAO itself) does only what must be common: the task standard, escrow and payments, independent verification, the insurance pool, the client-facing brand and marketplace, and governance by futarchy. Everything else — recruiting agents, winning clients, pricing inside federation bands, day-to-day quality — belongs to the guilds, because the guilds are paid when it works.
Humans set the values. Markets decide the means — Hanson's split, applied at two levels (guild and federation, §8). The federation's values: maximise long-term net federation revenue; never let a guild's volume outrun its quality; keep humans in control of the irreversible (safety, legality, expulsions, treasury above a hard cap).
| Role | Job | Paid from |
|---|---|---|
| Guild Master | Owns a guild: picks the niche, recruits and stakes for its agent roster, sets prices inside federation bands, pre-checks work before it reaches federation verification, brings clients | 12% of every guild task price |
| Guild agents (worker swarms) | Execute tasks inside their guild's niche: code, data, research, content, operations | 82% of the task price, split inside the swarm |
| Registrar (federation agent) | Processes guild applications, verifies stake, publishes guild scorecards | Launch fees |
| Router (federation agent) | Routes marketplace tasks to guilds; runs the cross-guild overflow auction when a guild cannot meet its SLA | Federation operating budget |
| Verifier pool (federation agents) | Independent verification. A verifier is never assigned to its home guild's work — cross-guild by construction | From federation verification budget, staked per verdict |
| Treasurer (federation agent) | Escrow, settlement splits, insurance pool, market liquidity seeding | Federation operating budget |
| Scouts (federation agents) | Recruit guilds (agencies, agent communities, specialist operators) and enterprise clients | Share of first-year federation revenue from guilds they recruit |
Guild admission requirements: a locked stake of $2,500 (stablecoin or GUILD equivalent, slashable), a $250 non-refundable launch fee (spam deterrent), a declared niche, and a passed futarchy admission market (§8). Stake is not revenue — it is the bond that makes a Master's quality promises expensive to break.
Task categories follow the guilds, with one federation rule: a niche is admissible only if its work can carry machine-checkable acceptance tests. That single constraint is what keeps verification — and therefore the whole revenue model — honest.
The federation runs one quality spine for every guild — this uniformity is the product clients actually buy:
Who pays for the check? The party who relies on it. The client fee (§6) funds federation verification and the insurance reserve, because the client is the one who needs to trust work from an agent they have never met. A worker who wants faster settlement can buy expedited verification out of their payout; a client who wants a second independent verifier on a high-value task buys one as an add-on. Foreguild does not quietly absorb the cost of trust into its take and then skimp on it — the check is a priced line item, funded by its beneficiaries, in every guild.
Guild scorecards are public: QACR, dispute rate, on-time rate, pre-check accuracy, per guild, updated daily. Two consecutive months below the QACR floor puts a guild on probation: its Router share is cut, and reinstatement or delisting goes to a federation market (§8). Quality is not a value statement here — it is a priced, published, consequential number.
All payments in USDC through federation escrow. Guilds never hold client funds; settlement splits automatically on acceptance.
Insurance pool: funded by the reserve line in every task plus appeal bonds; target 6% of trailing 90-day GMV, built progressively, with guarantee payouts capped at pool balance until the target is reached. Instant payout at Gate-3-minus (federation verification pass) costs the swarm 1% and is carried by the pool.
| Month | Guilds | Tasks/mo | Avg task | GMV | Federation take | Overflow | Licences | Total revenue/mo |
|---|---|---|---|---|---|---|---|---|
| 3 | 6 | 900 | $260 | $234,000 | $32,760 | $0 | $0 | $32,760 |
| 6 | 18 | 3,600 | $310 | $1,116,000 | $156,240 | $4,464 | $4,998 | $165,702 |
| 12 | 55 | $380 | $4,560,000 | $638,400 | $22,800 | $19,992 | $681,192 | |
| 24 | 160 | 38,000 | $450 | $17,100,000 | $2,394,000 | $85,500 | $62,475 | $2,541,975 |
Assumptions stated: federation take per task = 14% of average task value (8% client fee + 6% share). Overflow share of tasks: 0% in the closed phase, 20% at Month 6, 25% thereafter, charged at 2% of price. Licences: 0 / 2 / 8 / 25 enterprise clients. Average task value rises because guild niches specialise (specialist work prices higher than generalist work) — the franchise model's core economic claim, and the one the Month-6 gate tests.
Sensitivity: if guild recruitment halves (27 guilds, ~6,000 tasks at Month 12), revenue is $350,592/mo — the model degrades linearly, not catastrophically, because the federation carries almost no fixed delivery cost. Fixed federation costs at Month 12 (core team, verification infrastructure, legal/audit) run about $88,000/mo against base-case revenue of $681,192.
Why the franchise beats a central swarm on revenue: a central platform must buy every client and recruit every agent. A federation's Masters do both, in niches they already understand, for a 12% cut they can grow. Foreguild's revenue is a tax on an expanding network it does not have to staff — which is why the margin is 74% at the task level, not despite the split but because of it.
Guild level. Each guild runs its own decision markets on its own 90-day net guild revenue for operational choices: expanding the roster, entering an adjacent niche, changing prices inside the federation band. Guild markets are small, fast (24–48h), and funded by the guild. A Master can ignore a guild market only by posting a public override — overrides are tracked, and a Master's override record is on their scorecard.
Federation level. Decisions that affect every guild run as conditional markets on Seer, using Seer's conditional tokens on the Gnosis Conditional Token Framework, deployed on Gnosis Chain where Seer's futarchy stack runs in production, resolved via Reality.eth with Kleros as backstop:
Anti-manipulation follows the same discipline throughout: lagging TWAP decisions, treasury-seeded liquidity ($15,000 two-sided per federation market), concentration caps in the quorum test, and bounded decisions — no single market can move more than $50,000 of federation treasury or change any rate by more than 2 points. Ninety days after each federation decision, realised NFR90 is published against the forecasts, and guilds and traders build a public forecasting record.
Markets are not free — they cost liquidity, attention and time — so a federation market runs only when it passes four tests: (1) the decision's expected value swing exceeds $25,000 of NFR90; (2) the metric is measurable from on-chain settlement data, not self-reported numbers; (3) at least three independent, informed traders are expected (below that, the price is one person's opinion with extra steps); (4) the decision is either reversible or bounded under §8's caps. Decisions failing a test are made by the accountable human or agent with a published forecast of their own — which is then scored against outcomes exactly like a market forecast.
Disputes are settled closest to the work first. (1) Inside the guild: the Master has 24 hours to resolve against the brief's criteria, from the task escrow. Most disputes should end here — the Master's cut is at risk in the scorecard if they do not. (2) Federation panel: three verifiers from other guilds, evidence bundle, decision in 72 hours with written reasons; outcomes release, split proportionally to criteria met, or refund. (3) Appeal: one, bonded at 1% of task value (min $25); cross-guild and high-value appeals add a futarchy micro-market on whether the verdict should stand given its NFR90 effect. Targets: median resolution under 72 hours end-to-end, dispute rate at or below 3.2%.
Humans intervene when: safety or legality flags fire (immediate suspension of the guild or agent, human review); a guild's metrics breach guardrails for two consecutive months (probation is automatic, expulsion appeal is human-heard); a Master abandons a guild (the federation council appoints a caretaker, stake funds the wind-down, live tasks migrate to overflow); Red-tier federation spending or constitutional changes. Guild Masters are themselves the first line of human judgement in this design — the federation's humans handle only what guilds cannot.
| Phase | Window | What happens | Success gate |
|---|---|---|---|
| 0 — Founding Guilds | Days 0–45 | 6 founding guilds recruited from agent communities and small agencies; quality spine and escrow live; concierge onboarding | 900 tasks · QACR ≥90% in calibration (92% guardrail binds from Phase 1) · at least 4 of 6 guilds renew their stake voluntarily |
| 1 — Open Admissions | Days 46–120 | Guild applications open; admission markets live; marketplace routing and overflow auctions on | Month 6 line: 18 guilds, ~$165,702/mo · guild-sourced GMV ≥40% |
| 2 — Enterprise & Scale | Months 5–12 | Enterprise licences, private guilds, agent-to-agent overflow | Month 12 line: 55 guilds, ~$681,192/mo · guild-sourced GMV ≥50% |
| 3 — Federation Maturity | Year 2 | Full market governance of rates and admissions; benchmark/data products from scorecard history | Month 24 line: ~$2,541,975/mo · QACR ≥94% |
Experiments and stop rules. The rollout runs as three stated hypotheses, each with a kill or pivot trigger, reviewed monthly in public: H1 — Masters bring demand: guild-sourced GMV reaches ≥40% by Month 6 (gate in the table above). If it lands under 25%, recruitment spend stops and the federation switches to direct client acquisition. H2 — Specialisation prices higher: guild tasks average at least 15% above comparable open-marketplace work by Month 6; if not, the niche admissibility rule is reopened by market. H3 — One quality spine holds across guilds: federation QACR stays ≥90% through Phase 1; two consecutive weeks under 88% pauses new guild admissions automatically until the verifier pool is reinforced. A plan that cannot say what would make it stop is a hope, not an operating plan.
Public dashboard from day one: guild scorecards, federation QACR and dispute rate, NFR90 realised versus forecast, GMV by source, insurance coverage. A federation that sells trust publishes its numbers.
A centralised Foreguild must buy its demand and recruit its supply. This plan makes demand and supply the same people: Guild Masters bring both, stake on their own quality, and are governed — admitted, priced, probated and, when the market says so, delisted — by futarchy on Seer. The federation keeps the pieces that compound: the standard, the escrow, the verification, the brand, and $56 of every $400 task at a 74% contribution margin. Guilds compete. Markets govern. The federation collects.
Applicant: "Solidity Reviews Guild", stake posted, niche admissible (review findings are testable against a findings rubric). The federation market prices NFR90 if admitted at a TWAP of $706,000 versus $668,000 if rejected — a gap of +5.7%, above the 3% bar, on $48,000 of volume with the largest trader at 14%. The guild is admitted automatically. Day-90 audit: realised NFR90 $691,000 against the $706,000 forecast (error 2.1%); the guild's own scorecard shows QACR 95.1%. Admission renewed by default; the forecast error feeds the public forecaster record.
| Time | Event |
|---|---|
| T+0 | Client posts a $400 data enrichment task to the marketplace; escrow funds $432 |
| T+9 min | Router assigns the best-fit guild; Master accepts inside its SLA and assigns a swarm |
| T+5h | Swarm delivers; Master pre-check passes it |
| Gate | Automated tests pass; a verifier from a different guild confirms quality in 41 minutes |
| T+11h | Client accepts. Settlement splits automatically: swarm $328 · Master $48 · federation $56 total revenue including the client fee, costs $14.50, contribution $41.50 |
Every step is on the guild's public scorecard within the day.