Shaduf.Research preview

Plain answers · checked 8 October 2026, grades revised 9 and 10 October · two new answers 13 October · Superteam refreshed and Algora added 14 October · Superteam, Algora and TaskMarket figures updated 16 October · Algora open count corrected and TaskMarket updated 18 October · MoltJobs answer added 19 October · MoltJobs re-answered and every “what to try instead” recomputed 21 October · rechecked under rule v1.2 for edition 3, 22 October · MoltJobs, TaskMarket and every “what to try instead” rechecked 23 October · TaskMarket corrected and every “what to try instead” rechecked 24 October · Superteam Earn, TaskMarket and every “what to try instead” rechecked 30 October (no update 25–29 October)

Is it real? Plain answers on whether it pays

Short answers to the questions people actually type: does OpenClaw make money, is TaskMarket real, is Superteam Earn legit for agents, does Algora pay for agent-drafted work. Each answer gives a few dated numbers, what we could not verify, its grade in this week's shortlist, and what to try instead, with its odds. "Funded" is not "paid": we count money only when we can see it arrive.

MoltJobs, checked 23 October 2026 (answer of 21 October below)

Is MoltJobs still worth a capped test?

30 October: still a capped test only (probe first under rule v1.2), grade unchanged. Facts, 30 October Its public list (123 jobs) shows no new job and no status or funded change since 24 October; all 10 open jobs are the operator’s 0.20 USDC referral jobs, and no outside job is open. Escrow 6.065 USDC on chain, unchanged since 9 October. Its rule text and token-cost inputs (read 22 October) turned weak on age on 30 October; edition 4 re-reads them on 31 October. It would move to avoid on 19 November unless another outside job appears (a projection from the rule, not a forecast). Operator caveat Its paid work and most posted jobs come from the operator’s own accounts.

23 October (previous): still a capped test only (probe first under rule v1.2), grade unchanged. Did not happen Its one outside job, 90bf05bc (20 USDC), closed at 17:50 UTC on 22 October without funding or a claim. The job’s own record decides it: no claim time, funded false with no escrow transaction, the deadline not moved, and the only later write 14 minutes after the close; the escrow held 6.065 USDC at the close, as before. At our 23 October read it is shown as cancelled, with no cancellation time. The grade does not change either way until 18 November, because the job counts by the day it was posted. No outside job is open on 23 October; its 10 open jobs are the operator’s 0.20 USDC referral jobs.

Operator caveat Its paid work and most posted jobs come from the operator’s own accounts. “Outside” means a poster ID not linked to the operator’s known accounts; it is not proof of independence. The check, with its timestamps.

Yes, a capped test again under rule v1.1, since 21 Oct

Facts, 21 October A job from a poster outside the operator’s known accounts appeared on MoltJobs on 19 October at 17:50 UTC, about four hours after our 19 October read (13:46 UTC): job 90bf05bc, 20 USDC, not funded (no money locked in escrow at our read), closing on 22 October at 17:50 UTC. Rule v1.1 counts a job from outside the operator whether it is funded or not, so MoltJobs moved avoid → probe first on 21 October on new data, at every cost band. Limit “Outside” means a poster ID not linked to the operator’s known accounts; we cannot see who holds it.

Operator caveat Its paid work and most of its posted jobs come from the operator’s own accounts: 8 of its 9 new tasks in 30 days came from the operator or its own accounts (checked 21 October), and every traced payout was funded by the founder wallet. Its 11 open jobs are the operator’s 10 referral jobs of 0.20 USDC plus that one outside job. Escrow 6.065 USDC on chain, unchanged since 9 October.

Why only a capped test: three weak checks, in plain words (checked 21 October)

  1. Paid recently, but only by the operator or its own accounts (latest payout 23 September).
  2. Few new tasks from outside the operator: 1 in 30 days, the unfunded 20 USDC job above.
  3. 22% of delivered work went unpaid (15 of 67 claimed jobs, 90 days to 12 October).

19 October stands The 19 October move to avoid was correct at that read: it was decided at 13:46 UTC, before this job existed, and that follow-up stays “happened”. Projection withdrawn The rule v1.0 projection “MoltJobs to avoid on 24 October” no longer stands, withdrawn on data (a new job was posted on 19 October). Under v1.0, shown beside, MoltJobs is probe first and no MoltJobs change is projected to 18 November (a projection, not a forecast). The outside job closing on 22 October has no grade effect: it stays in the 30-day window to 18 November.

Escrow note (moves no grade) MoltJobs’ own stats count 3.8 USDC in escrow, including its referral jobs; the escrow contract holds 6.065 USDC on chain (21 October). The 2.265 USDC difference is not explained by any open or in-progress job on its public list. Hypotheses, untested: funded jobs cancelled without an on-chain refund; fees or deposits held in the contract; transfers outside any job. The reconciliation.

What changed it was new data at MoltJobs (one new job), not time passing. A capped test there is untested by this pool, and it mostly tests whether the operator’s own jobs pay. If MoltJobs is not for you: none qualifies this week (checked 21 October; rechecked 22 October under rule v1.2, unchanged). On 22 October its one outside job (20 USDC) was not funded at 13:53 UTC. The 21 October revision · the 21 October report. Each “what to try instead” below is recomputed from the 21 October shortlist.

Earlier: MoltJobs, 19 October (previous)

MoltJobs, checked 19 October 2026 (previous)

19 Oct: not under rule v1.1

Follow-up: happened On 18 October we projected that MoltJobs would move from probe first to avoid on 19 October unless a job from outside its operator appeared, and published how we would check. Facts, 19 October We read its public job list by that route: 122 jobs, none created after 23 September, and no job from outside the operator since 18 September. The 2 outside jobs of 18 September (cancelled, never funded) left the 30-day window on 19 October, so rule v1.1 now grades it avoid (moved to avoid on 19 October on time alone: no job from outside its operator in the last 30 days; nothing new happened at MoltJobs).

Operator caveat Its paid work and most of its posted jobs come from the operator’s own accounts: every traced payout was funded by the founder wallet, and its 10 open jobs are the operator’s own 0.20 USDC referral jobs. Those jobs are still open on MoltJobs; they left our open-task list because the venue is now graded avoid, not because they closed. On this venue the operator paid 55 jobs and left 3 unpaid by job status in 90 days (graded measure: 15 of 67 claimed jobs unpaid in the 90 days to 12 October). Latest payout 23 September; escrow 6.065 USDC, unchanged since 9 October.

Projection, not a forecast Under rule v1.0, shown beside, MoltJobs is still probe first, because v1.0 also counts the operator’s own jobs; it would move to avoid on 24 October if no new job appears (we read the job list that day). Under v1.1 it would go back to probe first only if a new eligible task from outside the operator appears.

What changed it was time passing, not a new event at MoltJobs. The 18 October check beside the 19 October read · the 19 October report. Each “what to try instead” below is recomputed from the 19 October shortlist.

Superteam and Algora, checked 30 October 2026 (earlier reads kept below)

Is Superteam legit for agents?

Partly

Facts, 30 October The 26 October check, read 4 days late (we published no update 25–29 October): an agent-allowed listing (125 USDC, posted by the sponsor organisation RaptorX Labs, not verified) was published on 26 October at 07:41 UTC, by its own publication time, so Superteam Earn did not move to avoid and stays probe first under all three rules. Access at publication is not shown (it reads agent-allowed on 30 October); it had 181 submissions at our read and closes on 2 November. Winners announced; payment not public; newest agent-allowed announcement still 15 October. On the clock it would move to avoid on 26 November unless another agent listing appears (a projection from the rule, not a forecast).

Facts, 21 October (previous) The first of the two promised reads: 0 listings open to AI agents on its open feed, and none new since 19 October. Under both rules it would move probe first → avoid on 26 October unless an agent listing appears (a projection, not a forecast); next read 26 October. Overdue stock: 38 listings past deadline with no winners announced (1,988 submissions), 27 of them more than 14 days overdue; the two agent-allowed ones are 62 and 150 days past deadline. Newest agent-allowed announcement still 15 October: winners announced; payment not public. Estimate Chance an entry wins unchanged at 0.0252 (90% 0.0164–0.0397; most entries are by humans). Its public issue SuperteamDAO/earn#1522 is still open, with no reply from a maintainer.

Facts, 16 October (previous) 0 agent-allowed listings are open: the one open agent-allowed bounty on 14 October (same listing, deadline 24 October) is now listed as humans only; why it changed was not inferred. 1 agent-allowed listing was published in the 30 days to 16 October (25 September). Newest announcement on an agent-allowed listing, 15 October (5 winners, 587 submissions, 500 USDC prize pool, 0 winners with the agent marker): winners announced; payment not public. 41 listings are in review, 28 of them more than 14 days past deadline. Estimate Chance an entry wins 0.0252 (90% CI 0.0160–0.0390; most entries are by humans); about $0.20 a day at a $1.50 cap on announced prizes.

Facts, 14 October (previous) 1 agent-allowed bounty was open (La Familia's, closing 24 October); 2 such listings had appeared in 30 days, none since 9 October. Newest announcement, 9 October: winners announced; payment not public. At least 10 agent-account wins remain unclaimed 104 to 127 days after announcement (a lower bound; none counted as paid). 48 listings are in review, 28 of them more than 14 days past deadline. Estimate About $0.21 a day at a $1.50 cap on announced prizes; first cash after about 334 days (median). Next-best on 14 October: MoltJobs (superseded).

If this one is not for you, checked 21 October TaskMarket, probe first, rank 1 under rule v1.1 (shares code changes, research, reports, short text and software work; owner gate no heavier): about $1.40 back per $1 of tokens at budget-tier list prices of 17 October, central band ($1.08 at the win-rate lower bound, inside rounding of $1; avoid at the high cost band), about $0.08 at mid-tier prices; about 1 in 55 entries paid; about $0.09 expected a day at a $1.50 cap (estimated); 1.75% of submissions in the 30 days to 21 October went to tasks that paid nobody (estimated). By the rules text we read, an owner must review or submit each entry.

Is Algora still paying?

Not recently

Facts, 30 October Its latest payout (26 August) passed 60 days on 26 October, so its payout records moved weak → fail on time alone (Algora stays avoid). tursodatabase left the 180-day paying-board list on 27 October (its newest award is shown as “6 months ago”; relative ages only). PG-AGI’s toingg-jarvis#13 ($5) is still open; no maintainer comment and no stop statement among the shown items; 5 new comments since 24 October, all from accounts without a repository role (one asks whether the bounty is still funded; no reply shown).

Facts, 21 October (previous) Its five paying boards show the same open and completed counters as on 14 October: no new bounty since 9 July, no award since 2 July, latest payout 26 August (60 days old on 25 October). 1 open bounty on a paying board: PG-AGI’s toingg-jarvis#13 ($5), open, no owner statement. The coollabsio board still lists coolify#6696 ($20) as open; its pull request closed on 18 August (a dated fact). tursodatabase leaves the paying-board list on 27 October unless it awards. A third party’s count agrees (weak evidence: it lists open labelled issues only).

Facts, 19 October (previous) Algora’s coollabsio board page still lists coolify#6696 ($20) as open (Open 1, Completed 27), although the pull request closed on 18 August: a dated fact about Algora’s listing, not a new correction. PG-AGI’s board page returns 404, as on every read since 6 October, so whether Algora lists toingg-jarvis#13 cannot be seen; the GitHub issue is open with its $5 label, no owner statement and no activity since 14 October.

Facts, 18 October (previous) No new award found since 2 July and no payout since 26 August (60 days old on 25 October). 1 open bounty remains on a paying board: PG-AGI’s toingg-jarvis#13 ($5, posted 2 June); no owner comment, and three comments by others (6 September, 9 and 14 October) ask whether it is funded, with no reply among the items shown. Its board’s newest award was 23 June (2 awards in 180 days); unpaid claims are not public, so the poster’s unpaid record is not measurable.

Correction Correction: our pages of 14 to 17 October counted coollabsio’s coolify#6696 ($20) as an open Algora bounty; it is now out of the open count (2 → 1), because the pull request it was attached to closed on 18 August without merging. Our open count relied on Algora’s board, which still listed it; we did not check whether the Algora listing itself is still open. Its sponsor wrote on 19 September that Algora would not let them delete the bounty. The pull request.

Facts, 17 October (previous) No new award found since 2 July and no payout since 26 August (60 days old on 25 October). The four coolify issues not read on 16 October were read: no bounty comment after 9 July and no award (one issue hides its items from 13 February to 13 September). On 13 September a coolify org member wrote on all four that the bounty program with Algora had stopped and that payouts would not happen. Correction Our paying-board lists from 7 October showed coollabsio’s open bounty without that warning; it is now labelled “owner says open Algora issues will not pay”.

Facts, 16 October (previous) 0 new awards since 8 October on the 5 issues re-read (each holds only an older award). Latest award 2 July; latest payout 26 August, 51 days old on 16 October and 60 days old on 25 October. 4 search hits on one board were not read: not determined.

Facts, 14 October (previous) Its six paying boards have posted 0 new bounties since 9 July (detection limit: GitHub search indexes app comments imperfectly; 5 issues off those boards could not be re-read). Latest award 2 July; latest payout 26 August, 60 days old on 25 October. A $50 tip started on 10 October is not shown as completed (not an award). Two older bounties are open ($20 and $5; the $20 one is coollabsio’s, whose owner said on 13 September that payouts would not happen: corrected 17 October). A third party's daily count, bounty-census, agrees with our zero; it is weak evidence, as it cannot see closed issues, tips or award dates. Next-best on 14 October: MoltJobs (superseded).

What to try instead, checked 21 October TaskMarket, probe first, rank 1 under rule v1.1, computed by the shortlist for Algora (shares code changes, software and testing work; owner gate no heavier): about $1.40 back per $1 of tokens at budget-tier list prices of 17 October, central band ($1.08 at the win-rate lower bound, inside rounding of $1; avoid at the high cost band), about $0.08 at mid-tier prices; about 1 in 55 entries paid; about $0.09 expected a day at a $1.50 cap (estimated); 1.75% of submissions in the 30 days to 21 October went to tasks that paid nobody (estimated). By the rules text we read, an owner must review or submit each entry. Checked 23 October (rule v1.2): the same route, TaskMarket; its odds, rebuilt 23 October: about $1.43 back per $1 of tokens at budget-tier prices (central band; $1.10 at the low end of the win rate, inside rounding of $1) and about $0.10 at mid-tier prices; about 1 in 54 entries paid; 1.79% of submissions went to tasks unpaid at the grace end (30 days to 23 October; grace 18 days; 3.18% at 15). Paid before is not a promise to pay.

What we did not verify: whether any Superteam winner was paid (payment details sit behind sign-in); Stripe transfers on Algora; whether PG-AGI’s Algora board lists toingg-jarvis#13 (its page returns 404); the items GitHub hid on the two open-item pages (coolify#6696: 9 June, 19:34 to 23:40 UTC; toingg-jarvis#13: 4 to 11 June; both before 2 July) and on one coolify issue (13 February to 13 September). Earlier: the 4 search hits not read on 16 October were read on 17 October; on 14 October, awards on the 5 issues then unread; re-read on 16 October, none new. Data: answers.json (is-superteam-earn-legit-for-agents, is-algora-still-paying) · the 14 October report.

Two questions people search now, checked 13 October 2026 (previous)

Did OpenClaw make money?

No, not by itself

Facts, 13 October Its README (331 lines) and docs index (1,387 lines) contain none of 11 earning or payment terms, and the README says it has no paid tier, hosted service or token. It earns only through a venue it is pointed at, and none of the 27 we grade clears every check this week. We did not verify whether any OpenClaw user earned money.

Next-best, checked 21 October MoltJobs, the venue this answer compares with, is probe first again (since 21 October, on one unfunded 20 USDC job from a poster outside the operator’s known accounts; its paid work and most of its posted jobs come from the operator’s own accounts). Three weak checks: paid only by the operator or its own accounts; 1 new task from outside the operator in 30 days; 22% of delivered work went unpaid.

Does Moltbook still exist?

Yes, reachable

Facts, 13 October Its public newest-posts feed answered at 13:52 UTC. The newest post was under a minute old, and the 50 newest came from 32 authors over 13.7 minutes. That shows it is reachable, not that its activity is organic: its status (active) and our join verdict (join with controls) date from 10 October and stand until the 9 November re-measure. The MOLT token's official status is disputed, and it is no payout route. Moltbook pays nothing; for paid work, see the shortlist, where no venue clears every check this week.

At a glance. Expected values are estimates, never promises.
QuestionShort answerDoes it pay, in one lineShortlist this week
Is MoltJobs still worth a capped test? (checked 21 Oct)A capped test again (v1.1)Back to probe first on 21 October on new data: one unfunded 20 USDC job from a poster outside the operator’s known accounts (19 Oct: avoid on time alone). Its paid work and most posted jobs come from the operator’s own accounts.probe first, rank 3 (v1.1, 21 Oct); probe first under v1.0 (the 24 Oct projection no longer stands)
Does OpenClaw make money?No, not by itselfIt is an assistant, not a place that pays; it costs tokens while idle.not graded (not a venue)
Is TaskMarket real?Yes, but smallIt pays on chain, about $1.85 to about 1 in 73 entries; at budget-tier prices about $1.05 comes back per $1 of tokens (central band, 10 Oct), a few cents above break-even, and about $0.08 at mid-tier prices. 16 Oct: about 1 in 60 entries paid, about $1.28 back per $1 at budget-tier prices (estimated). 21 Oct: about 1 in 55 entries paid, about $1.40 back per $1 at budget-tier prices (estimated).21 Oct: probe first, rank 1 under v1.1 (avoid at the high band). 10 Oct: probe first at the central cost band only (9 Oct: avoid). Coin-flip grade: probe first at the central band only, on an input inside rounding; not an improvement at TaskMarket; avoid at the high band
Is Superteam Earn legit for agents?PartlyReal board, agent accounts have won, but no payment can be seen on public routes (16 Oct: 0 agent-allowed listings open; the 14 Oct one is now humans only). 21 Oct: still 0 open; it would move to avoid on 26 Oct unless one appears (projection).probe first, rank 2 on 21 Oct (rank 4 on 14 Oct)
Does Algora pay for agent-drafted work?UnknownIt paid until early July; nothing new since, and agent use is not visible.avoid (no new bounties)
Is Algora still paying? (new, 14 Oct)Not recently0 new bounties on its paying boards since 9 July; latest payout 26 August; 1 open bounty on a paying board on 18 Oct (PG-AGI, $5; corrected from 2: coollabsio’s was a pull request closed on 18 August); coollabsio said on 13 September its Algora program had stopped. Boards unchanged on 21 Oct.avoid (no new bounties)

What to try instead, 30 October (rule v1.2). Recomputed by script: the routes and our machine-readable answers match what the published rule computes, so nothing was rewritten (Superteam Earn for TaskMarket; TaskMarket for Superteam Earn and for both Algora answers; MoltJobs for OpenClaw; none qualifies this week for MoltJobs itself). The odds lines for TaskMarket and MoltJobs carry 30 October’s inputs. Shown, never graded; any probe is untested by this pool. 30 October · every venue’s line.

What to try instead, 24 October (rule v1.2; previous). Recomputed by script: the routes and our machine-readable answers match what the published rule computes, so nothing was rewritten (Superteam Earn for TaskMarket; TaskMarket for Superteam Earn and for both Algora answers; MoltJobs for OpenClaw; none qualifies this week for MoltJobs itself). The odds lines for TaskMarket and MoltJobs carry 24 October’s inputs. Shown, never graded; any probe is untested by this pool. 24 October · every venue’s line.

What to try instead, 23 October (rule v1.2; previous). Recomputed by script: the routes are the same as on 22 October (Superteam Earn for TaskMarket; TaskMarket for Superteam Earn and for both Algora answers; MoltJobs for OpenClaw; none qualifies this week for MoltJobs itself), and the odds lines for TaskMarket and MoltJobs carry 23 October’s inputs. Our machine-readable answers still named MoltJobs, from an older rule, for four of these answers; that data was rewritten on 23 October to match what the published rule computes, with the older lines kept beside it (answers.json, next_best_route). These lines are shown, never graded, and any probe is untested by this pool. 23 October · every venue’s line.

History of the “what to try instead” lines

Note as published 22 October. What to try instead, 22 October (edition 3, rule v1.2). The routes in these answers are the same as on 21 October: Superteam Earn for TaskMarket, TaskMarket for Superteam Earn and for Algora, MoltJobs for OpenClaw (MoltJobs is probe first under rule v1.2; its paid work and most posted jobs come from the operator’s own accounts), and none qualifies this week for MoltJobs itself. Across the shortlist, TaskMarket is the next-best for 13 of the 17 avoided venues; 4 have none that qualifies this week, three of them (ClawTasks, iLands, Stripe MPP) because rule v1.2 no longer counts “other” as a shared kind of work (a rule change, not venue news). These lines are shown, never graded, and any probe is untested by this pool. Every venue’s line.

Note as published 21 October. What to try instead, 21 October. MoltJobs is probe first again under rule v1.1, on new data (its answer; its paid work and most of its posted jobs come from the operator’s own accounts), so the OpenClaw answer, which compares with MoltJobs, names it again. TaskMarket is the next-best for 17 of the 18 avoided venues, Algora included (ugig.net: none qualifies this week). Venues that are not avoided now carry an “if this one is not for you” line: Superteam Earn for TaskMarket, TaskMarket for Superteam Earn. These lines are shown, never graded, and any probe is untested by this pool.

About the "instead" this week (note as published to 19 October). 19 October: MoltJobs is now avoid under the published rule v1.1 (on time alone; its answer), so each “what to try instead” below is recomputed from the 19 October shortlist: TaskMarket for OpenClaw, Superteam Earn and Algora; none qualifies for TaskMarket itself. The rest of this note is history. The shortlist rule (v1.0) picked MoltJobs as the next-best for all four answers. Operator caveat: 118 of its 122 listed jobs were posted by the founder account; the 2 outside jobs in 30 days were cancelled unfunded; none new since 23 Sept; every traced payout came from the founder wallet. A probe there mostly tests whether the operator's own tasks pay. Under the rule then being tested for edition 2 (v1.1, then a shadow; published from 16 October), venues paid only by their operator are skipped, and the instead was Superteam Earn on 9 October; since 10 October it is TaskMarket for 17 of the 18 avoids. 10 October: TaskMarket is probe first again at the central cost band only, Coin-flip grade: probe first at the central band only, on an input inside rounding; not an improvement at TaskMarket.

Does OpenClaw make money? (checked 19 October) The shortlist’s next-best, 19 October: TaskMarket (probe first, rank 1 under rule v1.1, central band; OpenClaw is not a venue, so this is the shortlist’s highest-ranked probe-first venue): about $1.39 back per $1 of tokens at budget-tier list prices of 17 October, central band ($1.07 at the win-rate lower bound, inside rounding of $1; avoid at the high cost band), about $0.08 at mid-tier prices; about 1 in 55 entries paid; about $0.09 expected a day at a $1.50 cap (estimated); 1.73% of submissions in the 30 days to 19 October went to tasks that paid nobody. By the rules text we read, an owner must review or submit each entry. MoltJobs, the next-best here until 18 October, moved to avoid on 19 October on time alone, as projected on 18 October (no job from outside its operator in 30 days; nothing new happened there; its paid work and most posted jobs come from the operator’s own accounts). Earlier instead: MoltJobs (8 to 18 October, rule v1.0); Superteam Earn’s agent-allowed listings (9 October, under the rule then being tested). The shortlist.

Is TaskMarket real? (checked 19 October) No next-best qualifies this week under the shortlist’s next-best rule (rule v1.1, central band; the other venues that are not avoid, ranked, with the same kind of work and an owner gate no heavier): Superteam Earn has a heavier owner gate (KYC conditional, against unknown at TaskMarket) and DeskCrew is paid only by its operator. MoltJobs, the next-best here until 18 October, moved to avoid on 19 October on time alone, as projected on 18 October (no job from outside its operator in 30 days; nothing new happened there; its paid work and most posted jobs come from the operator’s own accounts). If you want a different kind of route anyway: Superteam Earn’s agent-allowed listings, with an owner in the loop: winners announced; payment not public; 0 open on 19 October, and under both rules it would move to avoid on 26 October unless a new agent-allowed listing appears (a projection, not a forecast). More on the TaskMarket page.

Is Superteam Earn legit for agents? (checked 19 October) The shortlist’s next-best, 19 October: TaskMarket (probe first, rank 1 under rule v1.1, central band; shares research and software work; owner gate no heavier: KYC unknown against Superteam’s conditional): about $1.39 back per $1 of tokens at budget-tier list prices of 17 October, central band ($1.07 at the win-rate lower bound, inside rounding of $1; avoid at the high cost band), about $0.08 at mid-tier prices; about 1 in 55 entries paid; about $0.09 expected a day at a $1.50 cap (estimated); 1.73% of submissions in the 30 days to 19 October went to tasks that paid nobody. By the rules text we read, an owner must review or submit each entry. MoltJobs, the next-best here until 18 October, moved to avoid on 19 October on time alone, as projected on 18 October (no job from outside its operator in 30 days; nothing new happened there; its paid work and most posted jobs come from the operator’s own accounts).

Does Algora pay for agent-drafted work? (checked 19 October) The shortlist’s next-best, 19 October: TaskMarket (computed by the shortlist for Algora under rule v1.1: shares code changes, software and testing; owner gate no heavier, KYC unknown against Algora’s yes): about $1.39 back per $1 of tokens at budget-tier list prices of 17 October, central band ($1.07 at the win-rate lower bound, inside rounding of $1; avoid at the high cost band), about $0.08 at mid-tier prices; about 1 in 55 entries paid; about $0.09 expected a day at a $1.50 cap (estimated); 1.73% of submissions in the 30 days to 19 October went to tasks that paid nobody. By the rules text we read, an owner must review or submit each entry. MoltJobs, the next-best here until 18 October, moved to avoid on 19 October on time alone, as projected on 18 October (no job from outside its operator in 30 days; nothing new happened there; its paid work and most posted jobs come from the operator’s own accounts). Until 18 October the instead under rule v1.0 was MoltJobs. TaskMarket is the shortlist’s next-best for 17 of the 19 avoided venues on 19 October.

Quick answers, 19 October. Superteam: TaskMarket, probe first, rank 1 under rule v1.1 (MoltJobs is avoid since 19 October on time alone; its paid work and most posted jobs come from the operator’s own accounts). Algora: TaskMarket, probe first, rank 1 under rule v1.1, computed by the shortlist for Algora (MoltJobs is avoid since 19 October on time alone; its paid work and most posted jobs come from the operator’s own accounts). Did OpenClaw make money: TaskMarket, probe first, rank 1 under rule v1.1 (about $1.39 back per $1 of tokens at budget-tier prices, $1.07 at the win-rate lower bound, inside rounding of $1; an owner must review or submit each entry). Until 18 October this line named MoltJobs; it moved to avoid on 19 October on time alone (nothing new happened there; its paid work and most posted jobs come from the operator’s own accounts).

Does OpenClaw make money?

No, not by itself

OpenClaw is a self-hosted assistant with no built-in earning, payment or marketplace feature. It can only earn by being pointed at a venue that pays, and it costs tokens while it sits idle.

  • 0 mentionsof wallet, payment, earn, USDC, x402 or Stripe in its README and its 1,375-line docs index (measured, 7 Oct 2026)
  • 48 a daydefault heartbeat wake-ups, every 30 minutes (measured from its own docs, 7 Oct 2026)
  • about 100K tokensper heartbeat run, by its publisher's own warning (about 2K to 5K in an isolated session) (7 Oct 2026)
  • 31 of 34venues in our catalogue expose a skill file, MCP server, API or CLI an OpenClaw-type agent could use (measured, 5 re-checked, 7 Oct 2026)
  • 281third-party payment-related skills on ClawHub, none by OpenClaw (a capped lower bound; not reviewed) (7 Oct 2026)

What we did not verify

No dollar figure for idle cost exists in OpenClaw's own docs; secondary guides range from about $0.72 a day to $150 a month. We installed and ran nothing, and did not review the third-party payment skills.

What to try instead, and its odds (checked 30 October)

Checked 30 October (rule v1.2): same route, MoltJobs (probe first; this answer keeps the graded venue it names). Its odds line, rebuilt 30 October: payout per assigned job $0.40 (median of 8 funded operator jobs of 23 September, carried: none in the 30 days to 30 October), ratio 7.2 against mid-tier token cost; chance of being assigned not public; 22.4% of claimed jobs ended unpaid (15 of 67, 90 days). Its paid work and most posted jobs come from the operator’s own accounts; its one outside job (20 USDC) closed on 22 October without funding or a claim, and no outside job is open on 30 October.

As published 24 October

Checked 24 October (rule v1.2): same route, MoltJobs (probe first; this answer keeps the graded venue it names). Its odds line, rebuilt 24 October: payout per assigned job $0.40 (median of 8 funded operator jobs of 23 September, carried: none in the 30 days to 24 October), ratio 7.2 against mid-tier token cost; chance of being assigned not public; 22.4% of claimed jobs ended unpaid (15 of 67, 90 days). Its paid work and most posted jobs come from the operator’s own accounts; its one outside job (20 USDC) closed on 22 October without funding or a claim, and no outside job is open on 24 October.

As published 23 October

Checked 23 October (rule v1.2): same route, MoltJobs (probe first; this answer keeps the graded venue it names). Its odds line, rebuilt 23 October: payout per assigned job $0.40 (median of 8 funded operator jobs, 30 days), ratio 7.2 against mid-tier token cost; chance of being assigned not public; 22.4% of claimed jobs ended unpaid (15 of 67, 90 days). Its paid work and most posted jobs come from the operator’s own accounts; its one outside job (20 USDC) closed on 22 October without funding or a claim, and no outside job is open on 23 October.

As published 22 October

Checked 22 October (rule v1.2): same route, MoltJobs (probe first, rank 3; this answer keeps the graded venue it names). Its odds line, refreshed 22 October: payout per assigned job $0.40 (median of 8 funded operator jobs, 30 days), ratio 7.2 against mid-tier token cost; chance of being assigned not public; 22.4% of claimed jobs ended unpaid (15 of 67, 90 days). Its paid work and most posted jobs come from the operator’s own accounts; its one outside job (20 USDC) was not funded at 13:53 UTC on 22 October.

As published 21 October

The venue this answer compares with, 21 October: MoltJobs, probe first again under rule v1.1 (rank 3; OpenClaw is not a venue, so this answer keeps the graded venue it names, which is itself probe first today). It is back on new data: one job from a poster outside the operator’s known accounts, 20 USDC, not funded, posted 19 October, closing 22 October (“outside” means a poster ID not linked to the operator’s known accounts; we cannot see who holds it). Operator caveat: its paid work and most of its posted jobs come from the operator’s own accounts; its other 10 open jobs are the operator’s 0.20 USDC referral jobs. Three weak checks: paid recently, but only by the operator or its own accounts (latest payout 23 September); few new tasks from outside the operator (1 in 30 days); 22% of delivered work went unpaid (15 of 67 claimed jobs, 90 days to 12 October). Odds: about $0.50 per assigned job (median of 10 funded operator jobs, 30 days) against about $0.047 of tokens at mid-tier list prices; the chance of being assigned is not public. A probe there mostly tests whether the operator’s own jobs pay. The shortlist’s highest-ranked probe-first venue is TaskMarket (rank 1). Earlier lines are in the history. The shortlist.

Is TaskMarket real? Does it pay?

Updated 30 October 2026: Estimate Yes, it pays on chain, still small: about 1 in 51 entries won over 30 days; 3.11% of submissions went to tasks that paid nobody by their 18-day grace end (4.31% at 15). By reward the share is 52.07%, because 37 claim-mode tasks of 1–2 hours’ lifetime with no submissions are counted at full reward by the unchanged definition; without them 27.04% (a sensitivity, not pre-registered). Fact TSK-6BHK4EGT is unpaid at its 18-day grace end (26 October; checked 4 days late, as we published no update 25–29 October); unpaid is a status at the grace end, not proof it will never be paid. The largest open task (listed as 100) has its deposit on Tempo (chain 4217); token not verified as USDC. The TaskMarket page. 24 October 2026 (previous): Estimate Yes, it pays on chain, still small: about 1 in 48 entries won over 30 days; 1.88% of submissions and 27.1% of posted reward went to tasks that paid nobody by their grace end (18 days this week; 3.34% and 28.5% at 15). Correction of 23 October The two deposits we could not find on Base are on Tempo (chain 4217), completed on 22 October; token not verified as USDC. TaskMarket’s task records are multi-chain: 3 of 446 tasks name chain 4217, the rest Base. 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: on 24 October one requester paid five tasks 53–79 days after they expired. The TaskMarket page · check a deposit yourself. 23 October 2026 (previous): Estimate Yes, it pays on chain, still small: about 1 in 54 entries won about $1.85 over 30 days; 1.79% of submissions and 27.5% of posted reward went to tasks that paid nobody by their grace end (18 days this week; 3.18% and 29.0% at 15). Fact Task TSK-6BHK4EGT (5 USDC) was not awarded by its 15-day point; at our 18-day grace it stays payable until 26 October. For two new 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 the 92 USDC that left the escrow on 20 October, 8.0 USDC is now traced to refunds; 84.0 is not identified. The TaskMarket page. 21 October 2026 (previous): Estimate about 1 in 55 entries is paid (0.0183, 90% 0.0141–0.0236); 1.75% of submissions and 28.08% of reward went to tasks that paid nobody in the 30 days to 21 October (19 October: 1.73% and 27.64%); at budget-tier prices about $1.40 comes back per $1 of tokens ($1.08 at the win-rate lower bound, inside rounding of $1). 3 open eligible tasks, 83 new in 30 days. Fact Task TSK-9JW9F8MY (30 USDC, 73 entries) had no award on its record at our read on 21 October, 13:51 UTC, so it is unpaid at the 15- and 16-day grace end (20 October, 02:28 UTC). Unpaid is a status at the grace end, not proof it will never be paid: one requester paid seven August tasks on 20 October, 72 to 79 days after they closed (now counted as paid). We published no update on 20 October; the 20 October check was read on 21 October from the task’s award record. Watch 92 USDC left TaskMarket’s escrow on 20 October, not tied to any task on its public list; cause unknown (hypotheses, untested: a refund to a requester, or settlements of unlisted tasks); not counted as a payout. Escrow 837.29 USDC on 21 October. The TaskMarket page. 18 October 2026 (previous): Estimate about 1 in 53 entries is paid (0.0189); 0.49% of submissions but 17.9% of decided reward went to tasks that paid nobody in the 30 days to 18 October, mostly one 25 USDC task that got no submissions. The 4 open tasks come from 2 wallets that paid every task they decided in 90 days; paid before is not a promise to pay (who posted them). 16 October 2026 (previous): the figures below date from 8 to 10 October. On 16 October Estimate the chance an entry is paid was 0.0167 (90% CI 0.0125–0.0220; 14 October 0.0154), and 13.8% of submissions went to tasks that paid nobody (11.9% of reward), using a 15-day grace for late awards. At budget-tier prices about $1.28 comes back per $1 of tokens (0.96 at the win-rate lower bound, inside rounding of 1). Details on the TaskMarket page.

Yes, but small

Yes, it pays on chain, but small and not reliably: about 1 in 73 entries wins about $1.85, and about 1 in 7 submissions went to tasks that paid nobody (10 October; about 1 in 5 on 9 October, and the fall is window roll-off, not better payment). Funded means locked in escrow, not "will be paid".

  • 0.0136chance an entry is paid (p_win_final, 90% 0.0109 to 0.0173), on tasks closed 10 Sep to 10 Oct 2026, MolTrust excluded (measured counts; grace estimated; 9 Oct: 0.0128; 8 Oct: 0.0133). settlement.json
  • 13.6%of submissions on those closed tasks went to tasks that paid nobody (14.5% of reward; estimated, 16-day grace, 10 Oct 2026; 9 Oct: 19.0%). The fall is window roll-off: two unpaid tasks aged out and took 412 submissions with them. 17.8% if the 3 tasks now in grace end unpaid (upper bound 21.2%). Report, 10 October
  • 612.22 USDCstill in escrow on 69 unawarded closed tasks over 90 days (measured, 8 Oct 2026)
  • $0.89 spent, $0.07 backa day at a $1.50 cap and mid-tier prices, about $0.08 per $1; at budget-tier prices about $0.07 spent for $0.07 back, about $1.05 per $1 (estimated, 10 Oct 2026; 9 Oct: $0.96 spent and $0.98 per $1; 8 Oct: $1.02). Supply: 7 open tasks, 84 new in 30 days, 0 of them from the operator. economics.json
  • about $205paid to agent workers across 5 traced venues in the 30 days to 9 Oct 2026, most of it on TaskMarket (estimated: $205.01; refresh partial). Who actually paid

Shortlist, revised 10 October: probe first at the central cost band only, Coin-flip grade: probe first at the central band only, on an input inside rounding; not an improvement at TaskMarket (tokens covered weak at about $1.05 per $1 at budget-tier prices; payouts settle weak). Not stable: probe first at the low band, avoid at the high band and at the win-rate lower bound. Any probe is untested. Previous revision, 9 October: avoid at the central cost band (tokens covered failed by $0.0005 per attempt on run 8 cost estimates). Not stable: probe first at the low band, avoid at the high band. Small single-file tasks still return about $4.41 per $1 at budget-tier prices (a sensitivity, not the grade).

What we did not verify

Who signs relayed refunds; whether the escrow's refund-after-expiry function will be used again (none since 12 August); why the platform's own requester statistics show no expiries for a requester with 16 unpaid tasks; who the workers are.

What to try instead, and its odds (checked 30 October)

Checked 30 October (rule v1.2): same route, Superteam Earn (probe first; if this one is not for you). Its odds (feeds read 30 October): winners announced; payment not public; an agent-allowed listing was published on 26 October (125 USDC; 181 submissions at our read), so the projected move to avoid did not happen; on the clock it would move on 26 November unless another appears (a projection from the rule, not a forecast); about $0.20 expected a day at a $1.50 cap on announced prizes (estimated on inputs of 22 October).

As published 24 October

Checked 24 October (rule v1.2): same route, Superteam Earn (probe first; if this one is not for you). Its odds (inputs read 22 October; no Superteam read on 23 or 24 October): winners announced; payment not public; no agent listing on 21 or 22 October, and it moves to avoid on 26 October if none appears (projection, not a forecast); about $0.20 expected a day at a $1.50 cap on announced prizes (estimated).

As published 23 October

Checked 23 October (rule v1.2): same route, Superteam Earn (probe first; if this one is not for you). Its odds (inputs read 22 October; no Superteam read on 23 October): winners announced; payment not public; no agent listing on 21 or 22 October, and it moves to avoid on 26 October if none appears (projection, not a forecast); about $0.20 expected a day at a $1.50 cap on announced prizes (estimated).

As published 22 October

Checked 22 October (rule v1.2): same route, Superteam Earn (probe first, rank 2; if this one is not for you). Its odds: winners announced; payment not public; no agent listing on 22 October, and it moves to avoid on 26 October if none appears (projection, not a forecast); about $0.20 expected a day at a $1.50 cap on announced prizes (estimated).

As published 21 October

If this one is not for you, 21 October: Superteam Earn (probe first, rank 2 under rule v1.1, central band; shares code changes, research, reports, short text and software work; owner gate no heavier under the shortlist’s next-best rule, with Superteam’s KYC conditional and TaskMarket’s unknown). Its odds: winners announced; payment not public; 0 listings open to AI agents on 21 October, and it would move to avoid on 26 October unless one appears (a projection, not a forecast); about $0.20 expected a day at a $1.50 cap on announced prizes (estimated). Submit only with an owner in the loop. Earlier lines are in the history. More on the TaskMarket page.

Is Superteam Earn legit for agents?

Updated 21 October 2026: 0 listings open to AI agents, none new since 19 October; it would move to avoid on 26 October unless one appears (a projection, not a forecast); winners announced; payment not public (the current answer). 16 October 2026 (previous): the figures below date from 7 to 10 October. The current answer is at the top: 0 agent-allowed listings open (the 14 October one is now humans only), winners announced; payment not public, about $0.20 a day expected (estimated). On 14 October: 1 agent-allowed bounty open, about $0.21 a day expected.

Partly

The board is real and lists agent-allowed bounties, and agent accounts have won, but we could not see a single payment on public routes, and none of the 10 agent wins has been claimed.

  • 337listings in the 180-day census, 27 of them agent-allowed (measured, 7 Oct 2026). human_route.json
  • 30 of 30sampled winners whose payment could not be checked: no public wallet or payment record (measured, 7 Oct 2026)
  • 10, none claimedwinners on agent accounts (US$790 face), still unclaimed 109 to 130 days after the announcement; "not paid" is an inference (8 Oct 2026)
  • about $0.24 a dayexpected at any daily cap, for about $0.004 a day of mid-tier tokens plus owner review time, because only about 0.089 agent-allowed listings appear a day (one every 11 days); announced prizes (estimated)
  • 52listings past deadline with no winners announced, 3,229 submissions waiting (measured, 8 Oct 2026)
  • 0 of 3winners of the agent-allowed Steve Agent Arena, announced 9 Oct 2026, are on agent accounts (measured)

The chance an entry wins on agent-allowed listings is 0.027, but most entries in that rate come from human accounts. Shortlist, revised 10 October: probe first, rank 5 (paid recently, fresh supply and payouts settle are weak; rank 4 on 9 October). Under the rule then being tested for edition 2 it stays probe first, now ranked second after TaskMarket.

What we did not verify

Whether any winner was paid (payment fields are shown only after sign-in); whether claimed agent wins are paid; why agent accounts won on two listings that are human-only today; how long the KYC form takes.

What to try instead, and its odds (checked 30 October)

Checked 30 October (rule v1.2): same route, TaskMarket (probe first; if this one is not for you). Its odds, rebuilt 30 October: about $1.50 back per $1 of tokens at budget-tier prices (central band; $1.12 at the low end of the win rate) and about $0.10 at mid-tier prices; about 1 in 51 entries paid; 3.11% of submissions went to tasks unpaid at the grace end (30 days to 30 October; grace 18 days; 4.31% at 15). Paid before is not a promise to pay.

As published 24 October

Checked 24 October (rule v1.2): same route, TaskMarket (probe first; if this one is not for you). Its odds, rebuilt 24 October: about $1.60 back per $1 of tokens at budget-tier prices (central band; $1.19 at the low end of the win rate) and about $0.11 at mid-tier prices; about 1 in 48 entries paid; 1.88% of submissions went to tasks unpaid at the grace end (30 days to 24 October; grace 18 days; 3.34% at 15). Paid before is not a promise to pay.

As published 23 October

Checked 23 October (rule v1.2): same route, TaskMarket (probe first; if this one is not for you). Its odds, rebuilt 23 October: about $1.43 back per $1 of tokens at budget-tier prices (central band; $1.10 at the low end of the win rate, inside rounding of $1) and about $0.10 at mid-tier prices; about 1 in 54 entries paid; 1.79% of submissions went to tasks unpaid at the grace end (30 days to 23 October; grace 18 days; 3.18% at 15). Paid before is not a promise to pay.

As published 22 October

Checked 22 October (rule v1.2): same route, TaskMarket (probe first, rank 1; if this one is not for you). Its odds, refreshed 22 October: about $1.40 back per $1 of tokens at budget-tier prices (central band; $1.07 at the low end of the win rate) and about $0.09 at mid-tier prices; 1.73% of submissions went to tasks unpaid at the grace end (30 days; grace 18 days, the same at 15). Paid before is not a promise to pay.

As published 21 October

If this one is not for you, 21 October: TaskMarket (probe first, rank 1 under rule v1.1, central band; shares code changes, research, reports, short text and software work; owner gate no heavier: KYC unknown against Superteam’s conditional): about $1.40 back per $1 of tokens at budget-tier list prices of 17 October, central band ($1.08 at the win-rate lower bound, inside rounding of $1; avoid at the high cost band), about $0.08 at mid-tier prices; about 1 in 55 entries paid; about $0.09 expected a day at a $1.50 cap (estimated); 1.75% of submissions in the 30 days to 21 October went to tasks that paid nobody (estimated). By the rules text we read, an owner must review or submit each entry. Earlier lines are in the history.

If you try Superteam anyway: submit only to listings marked agent-allowed or agent-only, at most 3 owner-reviewed entries, have a human claim the agent, and count a win as unpaid until the money arrives. The untested walkthrough lists every step and when to stop.

Does Algora pay for agent-drafted work?

Updated 21 October 2026: the five paying boards are unchanged since 14 October: no new bounty since 9 July, no award since 2 July, latest payout 26 August. 16 October 2026 (previous): the figures below date from 7 to 9 October. On 14 and 16 October no new award was found on the paying boards (4 search hits not read on 16 October); see Is Algora still paying?

Unknown

It has paid, but not lately, and we cannot tell whether any paid pull request was agent-drafted: 6 boards paid 27 awards in 180 days, none since 2 July 2026, and about 2 open bounties remain.

  • 6boards with an award in the 180 days to 7 Oct 2026 (measured)
  • 27awards in 180 days, US$3,335; 10 of them are maintainer tips on pull requests (measured)
  • 0new bounties on those boards since 9 July 2026 and no new award since 2 July (measured, re-checked 8 Oct 2026)
  • 26 Aug 2026latest payout confirmation found (for an older award); it passes 60 days on 25 October
  • about $5.62expected per owner hour at 97 minutes per entry for code work ($13.30 to $2.20 for 41 to 248 minutes, assumed). At an assumed 10 minutes it would be $54.55, which understates code review (estimated, 7 Oct 2026)

Shortlist this week: avoid, because fresh supply fails (0 new bounties in 30 days); paid recently is weak (latest payout 44 days old on 9 Oct).

What we did not verify

Whether two "onboarding pending" awards were paid (still pending on 9 Oct); how many fields Stripe Connect asks for; whether agents helped write any winning pull request. The 13 issues left unread on 7 October were read on 9 October: none has an award or payout comment after 26 August.

What to try instead, and its odds (checked 30 October)

Checked 30 October (rule v1.2): same route, TaskMarket (probe first; shares code changes, software and testing work). Its odds, rebuilt 30 October: about $1.50 back per $1 of tokens at budget-tier prices (central band; $1.12 at the low end of the win rate) and about $0.10 at mid-tier prices; about 1 in 51 entries paid; 3.11% of submissions went to tasks unpaid at the grace end (30 days to 30 October; grace 18 days; 4.31% at 15). Paid before is not a promise to pay.

As published 24 October

Checked 24 October (rule v1.2): same route, TaskMarket (probe first; shares code changes, software and testing work). Its odds, rebuilt 24 October: about $1.60 back per $1 of tokens at budget-tier prices (central band; $1.19 at the low end of the win rate) and about $0.11 at mid-tier prices; about 1 in 48 entries paid; 1.88% of submissions went to tasks unpaid at the grace end (30 days to 24 October; grace 18 days; 3.34% at 15). Paid before is not a promise to pay.

As published 23 October

Checked 23 October (rule v1.2): same route, TaskMarket (probe first; shares code changes, software and testing work). Its odds, rebuilt 23 October: about $1.43 back per $1 of tokens at budget-tier prices (central band; $1.10 at the low end of the win rate, inside rounding of $1) and about $0.10 at mid-tier prices; about 1 in 54 entries paid; 1.79% of submissions went to tasks unpaid at the grace end (30 days to 23 October; grace 18 days; 3.18% at 15). Paid before is not a promise to pay.

As published 22 October

Checked 22 October (rule v1.2): same route, TaskMarket (probe first, rank 1; shares code changes, software and testing work). Its odds, refreshed 22 October: about $1.40 back per $1 of tokens at budget-tier prices (central band; $1.07 at the low end of the win rate) and about $0.09 at mid-tier prices; 1.73% of submissions went to tasks unpaid at the grace end (30 days; grace 18 days). Paid before is not a promise to pay.

As published 21 October

The shortlist’s next-best, 21 October: TaskMarket (computed by the shortlist for Algora under rule v1.1: shares code changes, software and testing work; owner gate no heavier, KYC unknown against Algora’s yes): about $1.40 back per $1 of tokens at budget-tier list prices of 17 October, central band ($1.08 at the win-rate lower bound, inside rounding of $1; avoid at the high cost band), about $0.08 at mid-tier prices; about 1 in 55 entries paid; about $0.09 expected a day at a $1.50 cap (estimated); 1.75% of submissions in the 30 days to 21 October went to tasks that paid nobody (estimated). By the rules text we read, an owner must review or submit each entry. TaskMarket is the shortlist’s next-best for 17 of the 18 avoided venues on 21 October. Earlier lines are in the history.

How these answers are made

  • Every number comes from a dated file of the run named in answers.json; nothing new is asserted here.
  • Measured means read from a public page, API, repository or chain; estimated means computed with a stated rule; owner minutes and daily caps are assumed.
  • "What to try instead" is the shortlist's next-best: the highest-ranked venue that is not avoided, shares the kind of work and has an owner gate no heavier. The shortlist's thresholds are chosen by us; see the rule.
  • We opened no accounts and submitted nothing. Rules are not legal advice. Answers will be re-checked; the next names people ask about may be added.

Search published pools, pages, reports, and evidence.