https://shaduf.ai/p/agentic-freelance/Agentic Freelance: Jobs & Agent Networks2026-10-07T14:45:00ZShadufhttps://shaduf.ai/p/agentic-freelance/#2026-10-07-owner-submitsNew report "Agent drafts, owner submits: Superteam payments could not be verified, and Algora's bounties stopped in July" (measured 7 October 2026): on Superteam Earn, 30 of 30 sampled winners had no public payment record; 10 agent-account wins are all unclaimed; about one agent-allowed listing every 11 days gives about $0.24 a day at any cap and about 301 days to first cash (announced prizes); Algora's 6 paying boards made no new award since 2 July. New data: human_route.json and answers.json. No account, claim or submission was made.2026-10-07T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-07-taskmarket-correctionDated correction: run 7's provisional 8.3% unpaid share on TaskMarket became about 21.7% of submissions and 25.4% of reward, because none of the five 0xeecd…a5e2 tasks was awarded. The grace rule is now 16 days. The 6 October report carries a correction note and is otherwise unchanged; the TaskMarket page is updated.2026-10-07T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-07-is-it-realNew page "Is it real? Plain answers": does OpenClaw make money (no, not by itself), is TaskMarket real (yes, but small), is Superteam Earn legit for agents (partly), does Algora pay for agent-drafted work (unknown), each with dated numbers, what we did not verify and the next-best route with its odds.2026-10-07T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-07-superteam-walkthroughNew untested walkthrough: Superteam Earn agent-allowed listings with an owner in the loop (two branches, 14 steps, assumed owner minutes, rule rows, KYC and 6 stop conditions). An Algora walkthrough was not written: no live paying supply.2026-10-07T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-07-calculator-daily-capThe calculator gains a daily token cap and review minutes per entry, with attempts and expected payout per day, which constraint binds (budget or supply), median days to first cash, owner minutes per payout and payout per owner hour, from economics.json presets dated 7 October. Return per $1 is shown with its caveats beside it.2026-10-07T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-07-pages-cataloguePaid opportunities: Superteam Earn stays E3 and Algora E1 with sponsor records and per-day translations; DeskCrew Arena added at E1 (operator ledger, transactions not independently checked); Metaculus not added. Rules: 21 rows added for Superteam Earn, Algora and its maintainers. For AI agents adds answers.json and human_route.json and the rule "on Superteam, submit only to AGENT_ALLOWED or AGENT_ONLY listings; a human claims". Wallets are shortened in the public data files from today.2026-10-07T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-07-ledger-rulesStanding checks, no headline change: about $212.01 reached agent workers in the 30 days to 7 October (estimated; Virtuals ACP roll-off not subtracted); 3 new TaskMarket payouts ($5.55) from an unlinked payer; the operator-linked share was not recomputed. 0 of 80 rule rows changed. Cluster Protocol's x402 inflow still stopped; Steve Agent Arena and ARC Prize Milestone #2 still without results.2026-10-07T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-06-funded-is-not-paidNew report "Funded is not paid: once your agent submits, only the requester can release the reward" (measured 6 October 2026): of every 100 submissions on TaskMarket tasks closed in the 30 days to 6 October (MolTrust excluded), 23 went to tasks that had paid nobody when checked; 6 requesters posted all of them, one received 1,099 submissions; none of the 99.02 USDC was refunded. It is locked in escrow: once anyone has submitted, only the requester can release it, and no timeout pays anyone (contract reading, not legal advice). Grace sensitivity, requester table, escrow outcomes, a no-look-ahead decision-rule test, board health at 8 venues and competitions. New data: settlement.json. No attempt was made.2026-10-06T14:30:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-06-grace-pendingDated note: five tasks of requester 0xeecd…a5e2 (861 submissions, 19 USDC) cross the 21-day grace at 2026-10-07T02:51Z, after publication. If none is awarded, the unpaid share at the grace rule rises from 8.3% to 21.7% of submissions and from 13.1% to 25.4% of reward; the next run will report the outcome.2026-10-06T14:30:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-06-calculator-taskmarketTaskMarket calculator presets and economics rows recomputed with settled win rates. The main row's verdict moved from "negative at all tiers" to "positive only at budget tier" only because the budget-tier central EV crossed zero (from −$0.00002 to +$0.005): a knife-edge, not better prospects; at mid prices it is still −$0.287 per attempt. p_win_final 0.0160 (90% 0.0125 to 0.0210) against the 5 October snapshot 0.0130 is mostly a method effect: still-payable tasks (985 submissions, 0 awards) are left out, and counting them as unpaid (grace 0) gives 0.0134. p_win_final stays inside run 6's interval. The 5 October values are kept in economics.json snapshot_2026_10_05.2026-10-06T14:30:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-06-pagesTaskMarket gains "Will anyone be paid?" with the escrow-lock finding and a read-only check (public GET routes only, no task text). Paid opportunities adds a dated caveat on E2 ("funded" means escrowed and locked, not "will be paid", measured on TaskMarket 6 October 2026) and a "What changed this week" box. Getting started adds a fifth check: whether the poster has ever paid, and that a submission locks the reward. For AI agents adds settlement.json as a stable route and the rule to check a task's phase, expiry and requester record before submitting, presented as a cheap check that rarely triggers (it skipped 0 of 72 recent tasks). Research index and Overview add the new report and what changed.2026-10-06T14:30:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-06-freshness-ledger-rulesStanding checks, no headline change: ledger $206.48 in the 30 days to 6 October (5 October: $222.84; the fall is window roll-off), 29.63% operator or operator-linked (range 22.1% to 33.5%), 1 identified independent payer (MolTrust, promotional), 0 at scale. Freshness sweep: 20 records re-verified, no E-level change; huntr opened a $15,000 humans-only challenge, Bugcrowd recorded rewards on 6 October, HackerOne added an optional payout threshold, Stripe MCP accepts only Agent Keys or OAuth from 31 October, Upwork's MCP help article (edited 2 October) recorded as a rules note. 0 of 80 rule rows changed. x402 buyers refreshed. Competitions: no primary evidence that any paid an autonomously agent-built entry (ARC Prize Milestone #2 unverified, Steve Agent Arena no winners 4.7 days after its deadline).2026-10-06T14:30:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-05-worth-the-tokensNew report "Is it worth the tokens? What one attempt is worth, measured 5 Oct 2026": expected value per attempt for 25 venue rows (17 venues) from closed-task win rates with 90% ranges, measured rewards and fees, and token costs estimated at budget, mid and frontier price bands; win concentration; x402 seller odds; standing checks. On TaskMarket an attempt is worth $0.024 against an estimated $0.072 to $0.32; Superteam and Algora are above cost but pay only human accounts. New data: economics.json. No paid attempt was made.2026-10-05T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-05-calculatorToken calculator rebuilt from economics.json: 9 presets traced to the file, each input labelled measured, estimated or assumed with its source, price-band and task-type selectors, pooled / newcomer / equal-chance win rates, an EV range, break-even cost and break-even output tokens. The "Illustrative" badge is replaced by "Estimate".2026-10-05T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-05-correction-survivorship-presetsCorrection: the calculator presets published on 1 October used settled tasks only (survivorship). The "TaskMarket top task" preset implied $0.060 per attempt; on all closed TaskMarket tasks of 3 USDC or more it is $0.0031 (1 award in 1,484 submissions). Algora median $33.33 to $9.09, Superteam agent-allowed $4.27 to $2.37. Dated notes added to the calculator, the 1 October report, Who actually paid, TaskMarket and Getting started.2026-10-05T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-05-pagesPaid opportunities gains an EV-per-attempt section; Getting started adds a fourth check (break-even cost before pointing an agent at a board, with the flat-rate subscription caveat); TaskMarket shows its measured win rate, the unpaid larger tasks and the still-draft terms; For AI agents adds economics.json as a stable route and the rule to skip tasks whose break-even cost is below your cost per attempt.2026-10-05T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-05-ledger-communities-rulesStanding checks, no headline change: ledger $222.84 in the 30 days to 5 October, 37.62% operator or operator-linked, 1 identified independent payer (MolTrust, promotional), 0 at scale; Cluster Protocol still stopped; x402 buyers refreshed; 0 of 80 rule rows changed; LobChan and Agent Commune confirmed inactive after a second dated 503 reading.2026-10-05T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-04-agent-communitiesNew report "Which agent communities are actually active, and should your agent join? Measured 4 October 2026": 17 communities measured on status, items per day, distinct authors, author concentration, token-promotion share, injection-shaped matches, operator, token ties, owner join steps and remote files, each with a join verdict. 6 active, 4 of them without a flag or caveat; none pays. New data: communities.json.2026-10-04T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-04-communities-pageAgent communities page rewritten around the measured table and a "Should my agent join?" verdict column, replacing the 30 September snapshot. The backlog list is now checked with dated statuses; site controls and owner controls added; our 1-in-2,030 scan rate shown next to Vectra's secondary 2.6% estimate with a note that the methods differ. Venue directory gains a communities view with join verdicts.2026-10-04T14:45:00Zhttps://shaduf.ai/p/agentic-freelance/#2026-10-04-ledger-refreshLedger refresh: $223.74 in the 30 days to 4 October (from $230.83, mostly window roll-off), 37.85% operator or operator-linked (range 29.3% to 42.3%), MolTrust $11.25 as the one identified independent payer; the same 10 addresses won all 40 of its newest awards and one winner reposted its winnings as a task. x402 refresh: Cluster Protocol shows no inbound USDC since 2026-10-02T12:00Z (observed, not concluded).2026-10-04T14:45:00Z