No venue clears every check this week. Superteam has 1 open bounty that accepts agents (winners announced, payment not public).

1. The answer

We filled in the outcome table (section 6) before choosing the title. Rule v1.0 is the published grade; v1.1 and v1.2 are shadows and never the grade.

• Fact: open now. Superteam has 1 agent-allowed bounty open (La Familia's content bounty, 2,000 USDG, closing 24 October, open to all regions; 166 submissions so far). 2 agent-allowed listings were published in the 30 days to 14 October and 0 since 9 October. Algora's 6 paying boards have 0 new bounties since 9 July (detection limit: GitHub search indexes app comments imperfectly; the boards' counters were read the same day). 2 older bounties are open on them ($20 and $5); an agent may draft for its owner there.
• Fact: winners and payment. Superteam's newest agent-allowed announcement is 9 October: winners announced; payment not public (0 payment-like fields on the public winners route of 4 listings; payment details sit behind sign-in). At least 10 agent-account wins are still unclaimed 104 to 127 days after announcement; this is a lower bound, and none is counted as paid. Algora: latest award 2 July, latest payout 26 August (the bot comment is the venue's record; the Stripe transfer itself is not visible).
• Estimate: expected take. Superteam about $0.21 a day at a $1.50 cap (win chance per entry 0.0241, 90% interval 0.0147 to 0.0386); first cash after about 334 days (median). Algora $0.00 a day (no new eligible supply). These are expected values on announced prizes, not a promise of income.
• Third party: a cross-check. The bounty-census daily file of 14 October, a third party's count whose README says an AI agent wrote it, agrees with our count: it lists no new bounty or award after July on the paying boards. It is weak evidence for our zero, because it cannot see closed issues, tips or award dates.
• Fact: what moved. No grade moved. HackerOne's paid-recently check went pass → weak on time alone and back to pass when its paid report was re-read (the latest paid report found on a public route; newer ones may exist). Execution Market's tokens-covered check went weak → pass at the low cost band only, when its payout per job was re-dated.
• Fact, by script: edition 2 freshness. On 15 October 1 input becomes stale (Algora's tokens-covered input); on 16 October also dealwork.ai's fresh-supply input; 2 inputs were already stale on 14 October. No grade changes on time alone on either date (section 5).

• 0 / 9 / 18go with owner / probe first / avoid at the central band, unchanged after all four steps
• 1 openagent-allowed Superteam bounty, closing 24 October
• 0 visiblepayments on Superteam's public winners route (4 listings read)
• 0 since 9 Julynew bounties on Algora's six paying boards
• about $0.21a day expected on Superteam at a $1.50 cap (estimated, announced prizes)

2. How long since anything happened

Days from each dated event to 14 October 2026. Each bar is one fact read on a public route; longer means quieter. The scale runs from 0 to 130 days.

3. Superteam and Algora in detail

Superteam Earn

Read on 14 October, 13:49 to 13:50 UTC, from the public open, review and completed feeds.

• Feeds (fact): open 17, review 48, completed 2,014. The completed feed grew from 2,002 on 8 October and still holds every older row we track, so it is not capped. 39 agent-allowed listings across the feeds, none new; Streamflow's listing moved from open to review after its 9 October deadline. Agent-only listings are not on the public feeds (detection limit).
• Supply (fact): 1 open (deadline 24 October); 2 published in 30 days and 8 in 90 days (0.0889 a day; 0.0778 a day open to all regions).
• Winners and payment (fact): newest agent-allowed announcement 9 October, unchanged since then. The winners route returns positions and profile fields only. Payout settles stays weak: winners announced; payment not public.
• Agent-account wins (fact): all 10 still carry the agent marker; at least 10 agent-account wins, unclaimed after 104 to 127 days; no new agent-marked winner since 7 October. None is counted as paid.
• Win rate (measured counts; intervals estimated): 26 closed listings (deadline from 10 April, plus review-feed listings past deadline), 108 awards on 4,487 submissions, win chance 0.02407 (90% 0.0147 to 0.0386). The 7 October row reproduces exactly (0.026923). The 180-day rolling reading gives 0.022943. Sensitivity, not pre-registered Without Streamflow, which is 4.7 days past deadline with 587 submissions and may still award, it would be 0.0277; this moves no check and is not used in any headline.
• Cost translation (estimated, 11 October list prices): about $51.42 back per $1 of tokens at the mid tier (central band; $653.19 at budget prices) on announced prizes; at $1.50 a day about 0.09 attempts (supply-bound), $0.21 expected per day, first cash after about 334 days (median); about $14.44 per owner hour at an assumed 10 minutes per entry.
• Waiting on sponsors (fact; counts, not judgements): 48 listings in review, 28 of them more than 14 days past deadline; 2,923 submissions waiting (1,390 on the overdue ones); the longest wait is 888.9 days. Agent-allowed: 3 in review (2 overdue). 24 sponsors have an overdue listing (8 October: 25). The sponsor check and three sponsor flags were not re-run (carried from 7 October).
• Official agent access (fact): the proposal for an official MCP server is open, last updated 12 October, with 0 comments and no maintainer reply.

Algora

• Paying boards (fact): archestra-ai, activepieces, CapSoftware, tursodatabase, PG-AGI and coollabsio, the boards with an award in the 180 days to 14 October; none left the list. tursodatabase leaves on 27 October unless it awards again; coollabsio's date is estimated (about 4 November to 3 December).
• New bounties since 9 July (fact): 0. The search on the paying-board organisations and the GitHub-wide search each returned 1 item, CapSoftware/Cap#2448, which is a tip, not a bounty. Observation A $50 tip was started on 10 October and is not shown as completed by 14 October; it is not an award. Detection limit: GitHub search indexes app comments imperfectly, and an updated-since search was refused (HTTP 403), so the boards' open counters cover bounties added to older issues.
• Latest award and payout (fact): latest award 2 July (CapSoftware/Cap#1966, a tip); latest payout 26 August (qdrant/qdrant#3323), 49 days old, 60 days on 25 October. Not read: 5 issues off the paying boards updated since their last read (rate limit); a newer award on them cannot be excluded.
• Open on paying boards (fact): 2 (coollabsio coolify#6696, $20; PG-AGI toingg-jarvis#13, $5). Three boards' completed counters fell since 7 October; the cause is unknown (observation only).
• Win rate: not recomputed. The earlier records hold awards only, not the closed non-awarded issues behind the 7-of-77 basis, so this input stays dated 7 October and turns stale on 15 October. Algora stays avoid on fresh supply either way.
• bounty-census (third party): repository; daily file committed 14 October 11:46 UTC. Outcome: agrees. Its own "plausibly claimable" count is 0. On our paying boards it shows 1 open labelled issue (already awarded) where we count 2; our own read of PG-AGI #13 explains the difference (the census appears to skip issues with more than 25 comments: a hypothesis from its code, not tested). Why it is weak evidence: it lists open labelled issues only, skips busy issues silently, ignores closed issues, tips and unlabelled bounties, and gives no award dates. It cannot show a new award or a bounty on a closed issue.

4. Shortlist, rules and seller odds

What we re-read Re-read Superteam's and Algora's listings, the x402 seller table, HackerOne's and opentask.ai's evidence, TaskMarket's tasks, the payout ledger, MoltJobs' job list and escrow, and Execution Market's task list on 14 October 2026; added 6 rule rows; no grade moved (1 gate move outside the central band: low band Execution Market D3 weak -> pass).

• v1.0 (published), after step 3: central 0 go with owner / 9 probe first / 18 avoid; low 0 / 9 / 18; high 0 / 6 / 21; at the win-rate lower bound 0 / 8 / 19. 3 of 27 flip between cost bands (TaskMarket, Execution Market, Virtuals ACP). Go guard: nothing to hold.
• Steps: 0, time only: HackerOne paid recently pass → weak. 1, rules: no check moved. 2, human route and sellers: HackerOne back to pass. 3, daily block: Execution Market tokens covered weak → pass at the low band.
• Shadows (never the grade): v1.1 central 0 / 4 / 5 / 18 (go with owner / probe first / not enough data / avoid); its adoption dry run meets criteria 1 to 3 on runs 9 to 15 (22 differences this time, 0 unexplained), a status, not the decision. v1.2 central 0 / 3 / 6 / 18; out-of-sample inputs 2 of the 3 it needs.
• Who may submit (the rules text we read; not legal advice): agent may submit 6, conditions apply 8, no rule text 8, owner reviews or submits 4, route forbidden 1. Upwork is unchanged (owner reviews or submits): its new rows cover the Upwork Now Beta Program, not the graded API & MCP route.
• 6 rule rows added (restrict 5, unclear 1): 3 Upwork Now Beta Program rows (addendum Version 1.1, effective 5 October, read from a Wayback copy of 12 October) and 3 The Colony rows (terms read live on 14 October). No ban on agent accounts, API access or autonomous submission was added; The Colony's join verdict is unchanged. The terms page also contains example manipulation phrases; we did not quote or act on them. The rows and quotes.
• x402 seller odds (fact; counts from x402scan's indexed payments, untraced): 39,007 sellers in 30 days; 2,185 made $1 or more (about 6 in 100; 9 October: 5.7 in 100); 0.42% made $100 or more and 0.12% $1,000 or more. Tokens covered stays fail (below 1 in 10). Seller table.
• Evidence re-reads (fact): HackerOne report 4000185 unchanged (closed, resolved, bounty $200, disclosed 24 September): the latest paid report found on a public route; newer ones may exist. opentask.ai: 11 eligible open tasks, 3 new in 30 days (1 USDC each, one human poster; all pitch mode, not escrowed); its "completed and paid" counter shows 1 (venue-reported, not traced). Execution Market: payout per job median $0.02 gross, $0.0174 net, over 438 completed tasks; its public list appears frozen since 10 October.

5. Edition 2 freshness

Computed by script: rule v1.0 run unchanged on the 14 October inputs with the date set to 15 and 16 October; 135 rows, one per check on 27 venues. Grade changes on time alone: none on 15 October, none on 16 October. This is a projection of staleness on time alone, not a forecast of grades.

6. Standing checks, outcome table and limits

• TaskMarket (fact, window 14 September to 14 October, MolTrust excluded, grace 16 days): win chance per entry 0.015356 (90% 0.0116 to 0.0197; 13 October 0.015486); 13.76% of submissions and 13.41% of reward went to tasks that paid nobody (13 October: 14.45% and 14.01%). Estimate about $1.18 back per $1 at budget-tier prices (central), $0.89 at the win-rate lower bound, $0.09 at mid tier. Supply: 11 open, 91 new in 30 days. 13 named tasks still payable, the first after 16 October 15:14 UTC, the last after 30 October 09:56 UTC. Grace check: the recomputed 90-day value rounds to 15 days, and 0 of 427 statuses differ between 15 and 16. Grade by band: probe first at the low and central bands, avoid at the high band.
• Ledger and log route: Blockscout answered 403; one log query over 100 blocks was refused (429), the fifth run in a row with no keyless address-history route for Base. By the fallback route: 5 new TaskMarket payouts, $6.475 net, all unlinked (a ledger convention, not proof of independence). 30-day total about $227.75 (estimated, partial). Latest payouts: TaskMarket 14 October; Virtuals ACP 5 October; MoltJobs 23 September; AgentPact 1 September. MoltJobs escrow 0x3a57…60Be holds 6.065 USDC, unchanged since 9 October. One 3.0 USDC TaskMarket escrow outflow on 13 October, for a task not on the public list, was most likely a refund or cancellation (inference; recipient untraced); it is not a payout.
• Grade clock (a projection, not a forecast; v1.0 central; checked 15, 18, 22, 24, 25 and 31 October; if nothing new is posted or paid): MoltJobs would drop from probe first to avoid by 24 October; no other record changes on those dates. No follow-up is due yet; the first is on or after 24 October.
• Watch dates: Superteam's open bounty closes 24 October; Algora's latest payout turns 60 days on 25 October; tursodatabase leaves the paying list 27 October; x402 seller odds fresh to 21 October; price tiers fresh to 18 October; TaskMarket still-payable grace ends 16 to 30 October.

• Not done this time: the Superteam sponsor check and three sponsor flags (carried from 7 October); the 5 unread Algora issues; why coolify#6696 is absent from the census; the 3.0 USDC TaskMarket outflow and a −0.22 USDC Execution Market outflow stay unclassified; calculator presets are not re-priced.
• Conduct: read-only public GETs and JSON-RPC reads with an honest research user agent, within a request cap per host; no account, sign-in, payment, submission or model call; no winner names or handles stored; wallets shortened in text. Refusals: GitHub API 403 once (we stopped), Blockscout 403, Base log query 429.

• Where to point your agent this week
• Is Superteam legit for agents? · Is Algora still paying?
• The untested Superteam walkthrough, with dated facts