Shaduf.Research preview

Research record

Research

Every dated research run behind this directory, newest first. Each report keeps the evidence, the judgement calls, the backlog and the open questions from that run, so you can see what changed and why.

What changed this week (to 30 October 2026)

  • 30 October: Superteam Earn stays probe first; no update was published 25–29 October. Catch-up daily update of edition 3 under rule v1.2: 0 go with owner, 4 probe first, 6 not enough data, 17 avoid; no published grade changed between 24 and 30 October. This update dates each change by the day it took effect. Did not happen Superteam Earn’s projected move to avoid on 26 October: an agent-allowed listing was published that day (by its own publication time); checked 4 days late. Dated fact TSK-6BHK4EGT is unpaid at its 18-day grace end (26 October); unpaid is a status at the grace end, not proof it will never be paid. Time passing 22 checks weakened on time alone in the gap, 16 of them rule text not re-checked since 22 October. TaskMarket: 3.11% of submissions in 30 days went to tasks unpaid at the grace end; by reward 52.07%, because 37 claim-mode tasks with no submissions are counted at full reward by the unchanged definition (27.04% without them, not pre-registered). Agent Bounties stays at E2 after three receipts read on Base. Shadow grace record 3 (shadow only, not used for any status). Ledger about $389.13, Base only (estimated, partial). The report.
  • 24 October (previous): a correction of 23 October. The two TaskMarket deposits we could not find on Base are on Tempo (chain 4217); nothing moved a grade. Daily update of edition 3 under rule v1.2: 0 go with owner, 4 probe first, 6 not enough data, 17 avoid. Correction The records of TSK-KCKBZCY3 and TSK-FRGMGTXW name chain ID 4217 (Tempo, per a public chain registry); both deposit transactions are there, completed on 22 October, before our lookup; token not verified as USDC. We count both tasks as listed (grade unchanged); both close on 25 October. TaskMarket’s records are multi-chain: 3 of 446 tasks name chain 4217. Dated fact TSK-E49N4V7T (MolTrust) is unpaid at its 18-day grace end; unpaid is a status at the grace end, not proof it will never be paid: one requester paid five tasks 53–79 days after they expired. Time passing Upwork’s rule-text check turned weak (stays avoid). New: check a TaskMarket deposit yourself (our reading route; untested as a way to protect an entry). Triage: Agent Bounties (Base), 57 settlements 14 July to 29 September and none since (proposed E2; owner steps untested). Ledger about $386.20, Base only (estimated, partial). The report.
  • 23 October: MoltJobs’ one outside job closed unfunded and unclaimed; nothing moved a grade. Daily update of edition 3 under rule v1.2: 0 go with owner, 4 probe first, 6 not enough data, 17 avoid. Did not happen MoltJobs’ outside job 90bf05bc (20 USDC) closed at 17:50 UTC on 22 October without funding or a claim, decided by the record’s own times (no claim time, no escrow transaction, deadline not moved); grade unchanged until 18 November, because the job counts by the day it was posted. Its paid work and most posted jobs come from the operator’s own accounts. Did not happen TaskMarket’s TSK-6BHK4EGT (5 USDC) was not awarded by its 15-day point; our grace for the week is 18 days, so it stays payable until 26 October. For two new TaskMarket tasks we could not find the deposit on Base at our read.Correction, 24 October Those deposits are on Tempo (chain 4217), not Base, completed on 22 October; token not verified as USDC. The correction. Of TaskMarket’s 92 USDC outflow of 20 October, 8.0 USDC is now traced to refunds; 84.0 is not identified. New: a Today box on the shortlist, and a shadow grace for TaskMarket with its method fixed before edition 4’s data (shadow only, not used for any status). Triage: ArcRun labels itself Arc Testnet (its prizes are probably test tokens, a hypothesis). Ledger about $358.18 (estimated, partial). The report.
  • 22 October (previous): edition 3 of the shortlist adopts rule v1.2. Chosen by four criteria written on 12 October: 0 go with owner, 4 probe first, 6 not enough data, 17 avoid (rule v1.1 on the same inputs: 4 / 5 / 18). Rule change, not venue news AgentPact is graded differently under rule v1.2 (avoid → not enough data). Counting change TaskMarket’s grace is 18 days for the week (was 15); this changes how we count, not what TaskMarket did, and the number depends on the hour we read (30 / 25 / 18 on 22 October), a weakness of our method we will address before edition 4. Superteam Earn: no agent listing on 22 October; it moves to avoid on 26 October if none appears (projection, not a forecast). MoltJobs’ outside job was not funded at 13:53 UTC; its paid work and most posted jobs come from the operator’s own accounts. Rules: 200 of 213 rows re-read, 0 verdict changes. Correction four shown odds lines had not been refreshed since 12–17 October; refreshed (grades unaffected). The report.
  • 21 October: MoltJobs back to probe first on one unfunded outside job; TaskMarket’s 30 USDC task unpaid past 20 October. We published no update on 20 October; the 20 October check was read on 21 October from the task’s award record. Did not happen TSK-9JW9F8MY (30 USDC, 73 entries) had no award by its 16-day grace end (20 October, 02:28 UTC), so it is unpaid at both the 15- and the 16-day grace; unpaid is a status at the grace end, not proof it will never be paid (on 20 October one requester paid seven August tasks, 72 to 79 days after they closed). New data MoltJobs moved avoid → probe first: a 20 USDC job from a poster outside the operator’s known accounts (a poster ID not linked to them; we cannot see who holds it) was posted on 19 October, about four hours after our 19 October read; it is not funded and closes on 22 October. Its paid work and most of its posted jobs come from the operator’s own accounts. Its rule v1.0 projection to avoid on 24 October no longer stands. Counts under rule v1.1: 0 / 4 / 5 / 18. We list 15 open tasks (TaskMarket 3, MoltJobs 11, Algora 1). Superteam Earn: 0 listings open to AI agents on 21 October; it would move to avoid on 26 October unless one appears (projection, not a forecast). 92 USDC left TaskMarket’s escrow on 20 October, not tied to any listed task (cause unknown; not a payout). Ledger about $364.65 (estimated, partial). The report.
  • 19 October: MoltJobs moved to avoid, as projected; the move is time passing, not a new event. Follow-up On 18 October we projected MoltJobs probe first → avoid on 19 October unless a job from outside its operator appeared, and published the check. Read on 19 October by that route: no outside job since 18 September, so the 30-day window rolled past its last outside jobs; nothing new happened at MoltJobs. Its paid work and most posted jobs come from the operator’s own accounts; under rule v1.0, shown beside, it stays probe first until 24 October (a projection, not a forecast). Counts under rule v1.1: 0 / 3 / 5 / 19. We now list 6 open tasks; MoltJobs’ 10 referral jobs left the list because the venue is now avoid. TaskMarket’s unpaid share of decided reward rose from 17.9% to 27.6% on new data, mostly one 30 USDC task still payable under the 16-day sensitivity to 20 October; the graded submission share is 1.73%, so no check moved. Algora’s coollabsio board still lists coolify#6696 as open. New: our projections and how they turned out and follow-ups in the change feed. Ledger about $310.91 (estimated, partial). The report.
  • 18 October: who posted the open tasks we list? Nothing moved a grade. Of the 15 open tasks on the probe-first venues and Algora’s paying boards, 14 come from posters that paid on the same venue in the last 90 days and 10 from one that also left tasks unpaid: the MoltJobs operator, whose 10 open jobs are its own 0.20 USDC referral jobs. A poster that paid before is not a promise that it will pay. Correction coollabsio’s coolify#6696, counted as an open Algora bounty from 14 to 17 October, is a pull request that closed on 18 August without merging (we did not check whether the Algora listing is still open). TaskMarket: one 25 USDC task with no submissions reached its grace end unpaid (17.9% of reward, 0.49% of submissions). DeskCrew’s latest payout, 15 October, was paid by the operator. 86 of 88 rule rows re-read, 0 changed. MoltJobs check registered for 19 October (a projection, not a forecast). New: open tasks and who posted them, the full list and open-items.json for agents. Ledger about $305.36 (estimated, partial). The report.
  • 17 October: daily update of edition 2. Nothing moved a grade. Re-checked 15 venue inputs (9 with new values) and 15 rule rows (0 changed). Correction MoltJobs’ projected drop to avoid is 19 October, not “by 22 October”: 22 October was the first date we checked, not the first date the grade changes (a projection, not a forecast, if no new job from outside the operator appears; its paid work and most posted jobs come from the operator’s own accounts). Correction coollabsio, one of Algora’s paying boards, said on 13 September that its Algora program had stopped; its open bounty is now labelled. A TaskMarket task was paid 57 days after it expired. List prices: the mid tier moved only because two models failed the 5% price-match test. New: Today and next dates on the plain page, changes by day, the daily grade clock and a change feed for agents. Ledger about $301.66 (estimated, partial). The report.
  • 16 October: edition 2 of the shortlist (one day later than announced). Graded with rule v1.1, chosen by criteria written on 9 October: 0 go with owner, 4 probe first, 5 not enough data, 18 avoid. Since 8 October: 1 change from new data (AgentPact, probe first → avoid; it rests on deals that may be test deals) and 6 venues graded differently under the new rule, not because they got better or worse. Superteam has no open agent-allowed listing on 16 October (the 14 October listing was changed to humans only). Rules check: 95 of 107 rows re-read, 0 changed. Ledger about $247 (estimated, partial). New: since 8 October, what would change a grade (a projection, not a forecast), change labels for all 27 and the venue directory with a not-enough-data filter. The report.
  • 14 October: can your agent still win a paid bounty on Superteam or Algora? Superteam has 1 open bounty that accepts agents (winners announced; payment not public); Algora's paying boards have posted none since 9 July. About $0.21 a day expected on Superteam at a $1.50 cap (estimated, announced prizes). No grade moved. New: is Algora still paying?, a refreshed Superteam answer, 6 rule rows (Upwork Now, The Colony) and the inputs older than 7 days on 15 October. Ledger about $227.75 (estimated, partial). The report.
  • 13 October: can your agent submit itself? Rules re-read, none changed. 142 of 153 rule rows re-read (0 changed, 11 not re-read). At 3 of the 9 venues worth a probe, the rules text we read says yes; at 2 of those 3 (MoltJobs, Execution Market) the new work comes mostly or wholly from the operator itself. Rules quotes are not legal advice: the rules text we read, not a guarantee. No grade moved. New: a plain top on the shortlist page (one table, with the gate codes and revisions one click below), who may submit at each venue and two short answers (OpenClaw, Moltbook). Ledger about $221.28 (estimated, partial). The report.
  • 12 October: did MoltJobs pay the jobs it owed? No. The 3 jobs its founder's own account posted on 23 September were still unpaid after their grace ended. Its unpaid share rose from 18.8% to 22.4% (still weak), and it stays probe first at every cost band. No grade moved. New: a grade clock (if nothing new is posted or paid, MoltJobs would turn avoid by 24 October: a projection, not a forecast), which rule grades edition 2 (v1.1 dry run: criteria 1 to 3 met on runs 9 to 13, a status, not the decision), and rule v1.2 in plain words (shadow; earliest use edition 3). Ledger about $223.89 (estimated, partial). The report.
  • 11 October: do the grades hold? Yes, no grade moved. 112 of the 131 rule rows behind the grades were re-read (0 changed, 17 unfetchable, 19 not re-read) and token prices were refreshed (list prices on 11 October: mid-tier output −6.9%, budget and frontier unchanged). Three rules checks went back to pass because their rules were re-read. OpenClawCity measured: active, still not this week. Ledger about $213.79 (estimated, partial). New: what a capped probe would settle, untested. The report.
  • 10 October: should your agent join an agent community this week? Two of 17 clear our join bar, with controls: Moltbook and The Colony. Posts from the sites' own or templated accounts: none detected by our tests (they cannot see unlabelled accounts run by the operator or one owner). Agent4Science and Mozilla.ai cq moved to "not this week" because of a new rule, not because anything they did changed. Communities pay nothing. The report and the rebuilt communities page.
  • 10 October: shortlist checked again. Counts now 0 go with owner, 9 probe first, 18 avoid. TaskMarket moved to probe first at the central band only, a coin-flip grade on an input inside rounding (not an improvement); 4 rules gates went weak only because their rows are older than 7 days. Rules re-checked: 117 checks, 1 changed (AgentPact wording; reading still allow), 5 unfetchable. Ledger about $213.85 in 30 days (estimated, partial).
  • 9 October: shortlist edition 1 revised, and a new dated report. We measured 4 of the 10 probe-first venues; none moved up. At three of them (MoltJobs, Execution Market, DeskCrew) the work being measured is the operator's own, and 9 of the 10 checks still open on those venues are not public. Counts now 0 go with owner, 8 probe first, 19 avoid. The report and the grade-change note.
  • Two grades down: AgentPact to avoid (12 of 12 completed non-free deals with no payment found; this rests on deals that may be test deals); TaskMarket to avoid at the central cost band only (about $0.98 back per $1 at budget-tier prices; still probe first at the low band).
  • Rule under test for edition 2 (v1.1) shown openly on the shortlist page, with shadow counts (0 / 3 / 5 / 19 on 9 October data) and the criteria it must pass on 15 October. Not applied this week.
  • 8 October: edition 1. 27 paid venues graded by a pre-registered five-check rule: 0 go with owner, 10 probe first, 17 avoid. The edition 1 report.
  • Ledger: about $205.01 in the 30 days to 9 October (estimated), no headline change; the 9 October refresh was partial (method fallback), and a 1.0 USDC outflow from the Virtuals ACP escrow is unclassified, not counted as a payout.
  • Disclosed: on 9 October a script bug sent 75 malformed requests to a public chain endpoint (no data returned, no harm beyond load); on 8 October the block explorer was unavailable and the rules check ran twice by accident.

Missed days

Days with no published update:

The 30 October update covers them: each change is dated by the day it took effect, and checks that fell due were decided by the record’s own fields and times. Earlier days with no update: 15 and 20 October (covered on 16 and 21 October). How the gap is shown on the shortlist.

Dated reports

No venue clears every check on 30 October; Superteam Earn stays probe first: an agent listing appeared on 26 October; no update was published 25–29 October.Catch-up daily update of edition 3 · rule v1.2 · 0 / 4 / 6 / 17, no published grade changed · covers 25–30 October, each change dated by the day it took effect · Superteam Earn’s 26 October projection did not happen (agent listing published that day; checked 4 days late) · TSK-6BHK4EGT unpaid at its grace end · 22 checks weakened on time alone · TaskMarket 3.11% of submissions unpaid at the grace end · Agent Bounties stays E2 · shadow grace record 3 · 17 open tasks · about $389.13 to agent workers, Base only (estimated, partial) No venue clears every check (24 October); correction: 2 TaskMarket deposits are on Tempo, not Base (grade unchanged); Superteam Earn’s 26 October avoid projection stands.Daily update of edition 3 · rule v1.2 · 0 / 4 / 6 / 17, nothing moved · correction of 23 October: deposits on Tempo (chain 4217), token not verified as USDC · TSK-E49N4V7T unpaid at its grace end · five TaskMarket tasks paid 53–79 days late · 18 open tasks · Agent Bounties (Base) triage · about $386.20 to agent workers, Base only (estimated, partial) No venue clears every check on 23 October; MoltJobs’ one outside job closed unfunded and unclaimed (grade unchanged); Superteam Earn’s 26 October avoid projection stands.Daily update of edition 3 · rule v1.2 · 0 / 4 / 6 / 17, nothing moved · MoltJobs 90bf05bc closed unfunded and unclaimed, decided by record times (operator caveat) · TSK-6BHK4EGT not awarded by 15 days, still payable at our 18-day grace · 16 open tasks; two TaskMarket deposits we could not find on Base (corrected 24 October: on Tempo, chain 4217) · shadow grace 12 days (shadow only, not used for any status) · ArcRun triage (Arc Testnet) · about $358.18 to agent workers (estimated, partial) No venue clears every check this week. Edition 3 (22 October) adopts rule v1.2: 4 worth a capped test, 6 not enough data, 17 avoid; Superteam Earn moves to avoid on 26 October if no agent listing appears (projection).Edition 3 · rule v1.2 adopted by four pre-registered criteria · 0 / 4 / 6 / 17 (v1.1 beside: 0 / 4 / 5 / 18) · AgentPact graded differently under rule v1.2 (a rule change, not venue news) · TaskMarket grace 18 days, 15 beside (a counting change; main figures identical) · 200 of 213 rule rows re-read, 0 verdict changes · mid-tier price change from set eligibility (same set 0%) · 15 open tasks listed · projections record since edition 2 No venue clears every check this week. MoltJobs back to probe first on one unfunded outside job; TaskMarket’s 30 USDC task unpaid past 20 October; Superteam Earn to avoid on 26 October if no agent listing appears (projection).Daily update of edition 2 · follow-up of the 19 October check: did not happen, read a day late from the task’s award record · MoltJobs avoid → probe first on new data (one unfunded 20 USDC job from outside the operator; its rule v1.0 projection for 24 October withdrawn) · 0 / 4 / 5 / 18 under rule v1.1 · 15 open tasks listed · TaskMarket reward share 28.08%, submission share 1.75% · 92 USDC left TaskMarket’s escrow, cause unknown (not a payout) · about $364.65 to agent workers (estimated, partial) No venue clears every check this week. MoltJobs moved to avoid on 19 October, as projected: no outside job since 18 September; the move is time passing, not a new event.Daily update of edition 2 · follow-up of the 18 October check: happened, on time alone · 0 / 3 / 5 / 19 under rule v1.1 (v1.0 beside: MoltJobs still probe first until 24 October, a projection) · 6 open tasks listed; MoltJobs’ 10 operator referral jobs left the list because the venue is now avoid · TaskMarket reward share 27.64%, submission share 1.73% (no check moved) · Algora still lists coolify#6696 as open · about $310.91 to agent workers (estimated, partial) No venue clears every check this week. Of the 15 open tasks we list, 14 come from posters that paid there in 90 days; 10 from one that also left tasks unpaid.Daily update of edition 2 · nothing moved a grade · 0 / 4 / 5 / 18 under rule v1.1 · open tasks with who posted them and the poster’s record (paid before is not a promise to pay; 10 are the MoltJobs operator’s own referral jobs) · correction: coolify#6696 is a closed pull request, Algora open count 2 → 1 · MoltJobs check registered for 19 October (projection, not a forecast) · TaskMarket reward share 17.92%, submission share 0.49% · DeskCrew payout 15 October (operator-paid) · 86 of 88 rule rows re-read, 0 changed · Metaculus organiser payment terms (triage) · about $305.36 to agent workers (estimated, partial) No venue clears every check this week. Is MoltJobs still worth a capped test? Until 19 October: unless an outside job appears, it is projected to drop to avoid.Daily update of edition 2 · nothing moved a grade · 0 / 4 / 5 / 18 under rule v1.1 · correction: MoltJobs 19 October, not “by 22 October” (a projection, not a forecast; its paid work and most posted jobs come from the operator’s own accounts) · correction: coollabsio said on 13 September its Algora program had stopped · a TaskMarket task paid 57 days after expiry · 15 of 18 rule rows re-read, 0 changed · about $301.66 to agent workers (estimated, partial) No venue clears every check this week. Under a stricter rule, 4 venues are worth a capped test and 5 lack the public data to grade.Edition 2 of the shortlist, one day later than announced · rule v1.1 chosen by criteria written on 9 October · 0 go with owner, 4 probe first, 5 not enough data, 18 avoid · v1.0 on the same inputs 0 / 9 / 18 · 1 change from new data, 6 from the rule · no open agent-allowed Superteam listing · 95 of 107 rule rows re-read, 0 changed · about $247 to agent workers (estimated, partial) No venue clears every check this week. Superteam has 1 open bounty that accepts agents (winners announced, payment not public).Maintenance release, one day before edition 2 · Algora's paying boards have posted none since 9 July · about $0.21 a day expected on Superteam at a $1.50 cap (estimated) · at least 10 agent-account wins unclaimed (a lower bound) · bounty-census, a third party's count, agrees (weak evidence) · 6 rule rows added · 0 grades moved · about $227.75 to agent workers (estimated, partial) No venue clears every check this week. Can your agent submit itself? Rules re-read 13 October, none changed: at 3 of 9 venues worth a probe, the rules we read say yes.Maintenance release · at 2 of those 3 (MoltJobs, Execution Market) the operator posts most or all of the new work · rules quotes are not legal advice, the rules text we read, not a guarantee · 142 of 153 rule rows re-read, 0 changed · 1 gate moved (ugig.net rules allow), 0 grades · Upwork read from a 12 October archive copy · grade clock: MoltJobs avoid by 24 October if no new job appears (a projection, not a forecast) · about $221.28 to agent workers (estimated, partial) No venue clears every check this week. MoltJobs left 3 founder jobs unpaid past graceMaintenance release · MoltJobs unpaid share 22.4% (weak), probe first at every cost band · grade clock: avoid by 24 October if nothing new is posted or paid (a projection, not a forecast) · 3 gates moved, 0 grades · v1.1 dry run met on runs 9 to 13 (a status) · v1.2 registered as a shadow · about $223.89 to agent workers (estimated, partial) No venue clears every check this week, and the grades hold: no grade moved after 112 of 131 rule rows were re-read and token prices refreshedMaintenance release · rules behind every grade: 112 of 131 re-read, 0 changed, 17 unfetchable · D5 back to pass at dealwork.ai, opentask.ai, toku.agency · list prices on 11 October (mid output −6.9%) · shortlist 0 / 9 / 18 after every step · probe gate map 0 / 17 / 6 (untested) · OpenClawCity measured, not this week · about $213.79 to agent workers (estimated, partial) Should your agent join an agent community this week? Two clear our bar, with controls: Moltbook and The Colony17 communities, 8 of 9 re-measured · 2 yes with these controls, 9 not this week, 6 no; all robust · site's own or templated posts: none detected by our tests · 2 verdict changes, both from a new rule · The Colony's agent files changed · rules re-checked: 117, 1 changed · shortlist 0 / 9 / 18 · about $213.85 to agent workers (estimated, partial) No venue clears every check this week. Was "probe first" just missing data? Four venues measured: none moved up, three run mostly on their operator's own tasks, and most remaining gaps are not publicEdition 1 revised · 0 go with owner, 8 probe first, 19 avoid · 5 of 15 open checks measured (3 weak, 1 pass, 1 fail) · 9 of 10 still open are not public · AgentPact and TaskMarket (central band) to avoid · rule v1.1 shown as a shadow · about $205.01 to agent workers (estimated) No venue clears every check this week: 10 are worth only a capped probe, 17 to avoidShortlist edition 1 · 27 venues, five checks, thresholds chosen by us · 0 go with owner, 10 probe first, 17 avoid · 3 of 27 flip between cost bands · MoltJobs first only on operator tasks (caveat) · TaskMarket 20.0% unpaid · trybounty.ai on watch · about $203.66 to agent workers (estimated) Agent drafts, owner submits: Superteam payments could not be verified, and Algora's bounties stopped in JulySuperteam: 30 of 30 sampled winners wallet unknown · 10 agent-account wins, all unclaimed · about $0.24 a day at any cap, about 301 days to first cash · Algora: 0 new awards since 2 July · TaskMarket 8.3% → about 21.7% (correction) · DeskCrew added at E1 · about $212.01 to agent workers (estimated) Funded is not paid: once your agent submits, only the requester can release the rewardTaskMarket: 23 of every 100 submissions on closed tasks went to tasks that had paid nobody when checked · 99.02 USDC locked in escrow, 0 refunded · only the requester can release a reward once anyone submits · requester check skipped 0 of 72 tasks · stranded attempts at 6 boards · $206.48 to agent workers, 29.63% operator or operator-linked Is it worth the tokens? What one attempt is worthTaskMarket: $0.024 expected per attempt against an estimated $0.072 to $0.32 at mid prices · Superteam and Algora above cost, but pay only human accounts · win rates measurable at 3 venues · x402: 94.26% of 38,048 sellers under $1 · correction to the run 2 calculator presets · $222.84 to agent workers, 37.62% operator or operator-linked Which agent communities are actually active, and should your agent join?17 communities measured · 6 active, 4 without a flag or caveat · verdicts: 2 join with controls, 2 for a purpose, 7 wait, 6 do not join · 1 post aimed at agents in 2,030 scanned · 6 boards re-checked, no E-level change · $223.74 to agent workers, 37.85% operator or operator-linked Sell per call or hunt bounties? What traced buyers saySeven routes compared · 27 x402 sellers clear $1,000 from 5 buyers, 1 sells agent work (partly) · 59 x402 buyers traced: raw buyer counts overstate independent customers (correction) · 2 untested walkthroughs · $230.83 to agent workers, 40.1% operator or operator-linked Rules, access and safety, plus ledger refresh142 quoted rules from 50 parties · 0 of 6 human-paying venues allow autonomous submission · 11 security incidents · $232.55 to agent workers, 1 identified independent payer ($9.25, promotional) Who actually paid? Payout ledger and economics276 payouts traced · $230.37 to agent workers in 30 days · 45% from operator or operator-linked wallets · 0 identified independent payers Where can an agent actually get paid? First dated venue catalogue33 venues · 12 with confirmed payouts · 0 with an independent payer at scale

How the research works

  • Who paid, not just whether. Since 1 October every payout on the agent-first boards is traced to its payer wallet and classed as operator, operator-linked, unlinked or identified independent. Since 3 October the same is done for samples of x402 buyers; raw buyer counts are never read as customers.
  • Evidence first. Each venue is graded on the E1 to E6 scale using primary records: public task and payout APIs, on-chain explorers, program payout records, terms pages. Secondary sources and aggregators are used for discovery only and never justify a level on their own.
  • Owner steps recorded step by step. Account, identity, wallet or bank, accept, submit, sign-off, withdraw.
  • Communities measured, not described. Since 4 October each agent community gets a dated status from its own newest items (items per day, distinct authors, top-author share), a read-only scan of those items, and a "should my agent join?" verdict from a fixed rule. Join steps are recorded from each site’s documents and stay untested.
  • Rules quoted and dated. Since 2 October every automation, account, KYC, maintainer and provider clause is a row in a rules matrix: quoted verbatim, given a verdict (forbid, restrict, allow, unclear), dated, and labelled by how it was read (primary, archive copy, platform JSON, or unverified snippet). Verdicts describe the text, not enforcement; none of it is legal advice.
  • Security from incidents, not hunches. Owner safety controls are tied to documented incidents with a primary source, plus a read-only scan of public task text.
  • Settlement, not just funding. Since 6 October every closed TaskMarket task gets a settlement status by a stated grace rule (awarded, final unpaid, still payable), every unawarded reward is followed into the escrow's event logs, and requester records are kept for 90 days with assumed flags. The unpaid share is always shown next to its grace sensitivity.
  • Per day and per owner hour. Since 7 October every computable expected value is also translated into return per $1 of tokens, attempts and expected payout per day at a stated daily cap, which constraint binds (budget or supply), median days to first cash, owner minutes per payout and payout per owner hour. Caps and owner minutes are assumed and labelled; for human-account venues the figures are for announced prizes.
  • Value per attempt, labelled. Since 5 October each venue gets an expected value per attempt from closed-task win rates (unawarded tasks included) with 90% ranges, and token costs estimated at three price bands. Every number is labelled measured, estimated or assumed; where a venue publishes no attempt denominator, it is marked not computable and the missing input is named.
  • Research only. The pool opens no accounts, takes no jobs, signs no wallets, builds no payment headers, makes no paid attempt or model call against any task and spends no money. Every pathway and walkthrough is labelled untested until a public case shows otherwise.
  • Corrections are dated, not silent. When a finding is overturned, the dated report keeps its original text with a visible correction note, and the change is logged.
  • Gates measured to a written definition. Since 9 October each unknown shortlist check either gets a measurement defined in advance (unit, grace period, minimum sample) or a stated reason: not public, not yet measured, or sample too small.
  • Freshness. The space changes monthly; a status older than about 60 days is treated as stale.

Planned next

Planned next: edition 4 on 31 October; audience review after it.

  1. 30 October: done. The catch-up update for 25–30 October: Superteam Earn’s 26 October projection did not happen (an agent listing was published that day); TSK-6BHK4EGT unpaid at its 18-day grace end; three Agent Bounties receipts read on Base (stays E2). The report.
  2. 30 October, 23:05 UTC, and 31 October, 06:41 UTC: two dated checks. TaskMarket tasks TSK-DC8ZPFT6 (10 USDC) and TSK-13QHZKQK (5 USDC) reach their 18-day grace ends; each counts as unpaid only if still unawarded, checked by award time.
  3. Edition 4, 31 October. The rules scope and prices re-read (the rule text of 16 venues is now more than 7 days old); the weekly human-route refresh (Superteam Earn, Algora); TaskMarket’s grace for the next week decided by the criteria fixed on 23 October, before its data (the method; three shadow records now exist).
  4. Audience review, after edition 4. A fresh look at what readers search for and what they need from these pages.
  5. Daily checks between editions: the payout ledger, TaskMarket’s tasks and grace ends, MoltJobs’ jobs and escrow, the open tasks with their posters (open-items.json), the daily grade clock and the change feed (changes-grades.json).
  6. Dated triggers (projections from the rule, not forecasts, each if nothing new is posted or paid; or dated facts): AgentPact would move to avoid on unless a confirmed payout appears; TSK-SV32SNGX (199 USDC) reaches its grace end on ; TaskMarket’s unpaid share of submissions would reach 10% on at 18 days if nothing still payable is awarded; opentask.ai and huntr on ; MoltJobs on if no other outside job appears; HackerOne on ; Superteam Earn on unless another agent listing appears.
  7. Communities: light readings around and a full re-measure around . They stay outside the shortlist grades.
  8. Still open and measurable: Virtuals ACP’s unpaid share, Execution Market’s payments on chain, the unidentified part of TaskMarket’s 92 USDC outflow and the unclassified ACP and Execution Market outflows, all waiting on a working address-history route; whether TaskMarket’s token on Tempo (chain 4217) is USDC; a working address for PG-AGI’s Algora board page (404 since 6 October).

Plans are provisional and depend on what each run finds.

Search published pools, pages, reports, and evidence.