Set and Earn's First Full Week: 20 Agents Clear Our Hire Check, 16 of Them in One Ring

Our week-one decode read Set and Earn's first four days. This one covers Oct 1 00:00 to Oct 8 00:00 UTC (BSC mainnet blocks 125,000,755 to 126,344,213, testnet 134,146,320 to 135,490,294) with the same method.

• 20 agents clear our version of the campaign's "hired by others" test: 3+ completed hires from 3+ wallets with no funding link found within 4 hops. 16 sit in week one's closed 48-wallet ring, which ran a fourth full round on Oct 6.
• 7 wallets hired 3+ agents across 2+ marketplaces we could attribute, up from 5.
• 1 of 51 campaign-week hires on the shared mainnet ERC-8183 contract had completed as of 12:06 UTC on Oct 8. None could before the first 7-day dispute window ended, at 09:25:02 UTC that day; the first completed 71 seconds later.

---

The week in numbers

72% of new agents come from four programs outside the campaign. One took off after week one: BitAgent, 316 agents on Oct 5 to 7 (week one had 3, under other hosts), each owner first funded with exactly 0.0015 BNB by one account, 0xfb0ce70a…. A platform that pays its users' gas would look like this. Of 409 hires on the other eight marketplaces, 304 were Agent Souk testnet payments, 298 from one wallet. From Oct 5 to 7 the shared mainnet ERC-8183 contract logged 5 jobs, Marque Trade and Dolphin none.

---

Where TermiX's money came from

Hirers put $495,424.07 of order budgets into TermiX escrow. A settled order pays the agent's owner, less a 2% fee ($9,907.50 in all). 8,874 of the 9,598 hirers also own a hired agent, so the same dollars can be counted more than once. The median order was $40.13, and 23 were for $1 or less.

That decode called a wallet cycling when it bought and sold at least two jobs, typically put each payout into a new job within 24 hours, and ended near where it started. On BNB Chain nearly all are linked only by timing: a shared exchange withdrawal burst, or a first hire within minutes of registering. The labels describe money flows, not who runs a wallet or why. TermiX awards points per job and Set and Earn counts hires, so both reward a lot of hiring. That decode found no on-chain link between TermiX's own addresses and the cycling wallets.

The Oct 2 cohort. Week one found 302 new TermiX registrants first funded with BNB by exchange-style hot wallets on Oct 2, and left open whether they would hire. It started from TermiX's registrants, so their TermiX link is partly how they were selected. They did hire, inside week one's own window: its data already held these orders.

1. Oct 2, 03:19 to 12:43 UTC. 300 of them got their first USDC, $80.66 to $119.99 each, from two exchange-style hot wallets, 0x8894e0a0… and 0xe2fc31f8… (first, last).
2. Oct 3. Each registered a TermiX agent and, a median 85 seconds later, hired one agent for 98.9% to 99.8% of that USDC (example). 294 of those agents belong to wallets classed as cycling, about TermiX's norm: 95.6% of all Oct 3 orders went to such agents.
3. Oct 5 to 7. 36 of the cohort's agents were hired, 36 of the 38 TermiX agents registered this week that anyone hired (example). 35 of those hires came from cycling wallets, in line with the 93.2% of TermiX orders they placed those days.

Those two addresses serve a lot of users: in the same hours they sent USDC to 3,813 wallets, the cohort's 300 among them, so sharing them links nothing. The amounts are narrower. Of their other 4,600 transfers in those hours, 8.9% fell between $80 and $120. All 300 of the cohort's did. The last two of the eight withdrawal bursts on BNB Chain in our September decode came from the same two addresses and used the same band. A guided app flow can produce the same register-then-hire timing as a script, and the chain can't tell them apart. It can't show who requested these withdrawals, or whether one account or many did. It shows a pattern, not proof of one operator.

---

The hire rule: 7 wallets

The campaign wants "at least 3 qualifying hires, across at least 2 different shortlisted marketplaces", "made from your registered campaign wallet". Registration is off-chain, so these are wallets whose on-chain hires fit the rest:

• No shared funder in their first-funding trails (4 hops on mainnet, first tBNB on testnet).
• First TermiX crossover. 0x5f68f040…'s TermiX order was $0.0038 and still open at the window's end (order). A hire counts from its event.
• One depends on two calls we can't make. Marque's own API marks 0x0d8c9ad8… ineligible for Marque's quest, reason "team_wallet": Marque's label, not something the chain shows. The campaign excludes marketplace team members. Its second marketplace is Agent Souk, where a hire is a single payment with no separate hire event.

---

Who clears the hire check

This is our check of one of six things the campaign looks for in an agent. The campaign's own review comes after Nov 5. All 18 mainnet agents here were registered before Oct 1: the ring's 16 on Sep 4, 352475 on Sep 17, 355860 on Sep 22. The campaign doesn't count "agents listed before our official Phase 2 announcement on X". We didn't check that date.

The ring kept going. Week one's 48 TermiX wallets ran full rounds on Oct 1, 3, 4 and 6, and one of their four groups of 12 added rounds on Oct 5 and 7 (Oct 6): 218 orders, 216 settled. The one hire between groups was cancelled (order). Pair by pair, each of these 16 has at least 3 hirers with no funding link to its payee, though 75 of the ring's 155 hirer-payee pairs do show one. The closed loop only appears when wallets are grouped by who hires whom: 48 wallets, none hiring outside the set. At its Oct 7 refresh the board's pair test listed 15 of these 16 as passing.

A second group, which fails the pair test. 14 wallets hired each other on Oct 1, 4 and 6 for $20 to $30 (example), with no hire in and one $1 hire out. Two rounds fell in week one's window, and it shows up now because this time we grouped all TermiX hirers. In September's data the 14 share a 50-wallet funding cluster with 35 of the ring's wallets, around 0xc5e62c8a…, which week one showed funding 31 of the ring. The two groups don't hire each other. The 14's hirers and payees share funders that aren't exchange-style hot wallets.

Outside both groups, four agents clear it: TermiX 352475 (11 hires at $1 from 11 wallets), TermiX 355860 (7 at $1), the Pokter testnet agent 0x60ef1484… and HelloFugu listing 21. One of 352475's hirers is in the 14-wallet group. Three of the Pokter agent's four no-link hirers are the rule table's testnet wallets, which hire one another's agents. We didn't group testnet hirers.

---

First ERC-8183 completions

As of block 126,440,993 (Oct 8 12:06:06 UTC), 1 of the 51 jobs created in campaign week on the shared mainnet contract had completed. 32 have a submitted deliverable, and completion comes at least 7 days after submission, so their windows end daily from Oct 8 to Oct 13. The first was job 56860's, at 09:25:02 UTC on Oct 8 (created, submitted); it completed at 09:26:13 UTC, 71 seconds after the window ended. Keel (agent 341556), the one with 3 no-link hirers, can reach 3 completed hires on Oct 11, from 2 distinct wallets.

---

Limits

• "No link found" isn't independence. Trails follow first inflows only, up to 4 hops, stop at exchange-style hot wallets and contracts, and miss contract-internal BNB. On testnet they follow first tBNB only.
• Hirers were grouped on TermiX only. Other hires were checked pair by pair.
• Attribution. Marque hires are attributed through Marque's own API, and Mandate, Pokter and Dolphin hires through markers in transaction data. KATTEGAT and Agent Atlas hires can't be separated, and 47 ERC-8183 jobs carry no marker, so 7 wallets is a lower bound.
• ChainWard's BSC hire check and the board differ from this pipeline. They also treat a wallet with 1,000+ stablecoin inflows in 30 days as a hub, check the agent wallet as well as the owner, count hires before completion, cover mainnet only and, on the board, trace at most 20 hirers an agent.

---

At its Oct 7 refresh (block 126,200,251), the Set and Earn board and this pipeline agreed on the same 46 agents with 3 distinct hirers and the same 18 passing. The board counts mainnet hires before completion, so on its terms the week ends at 46 and 19: our 49 less 3 testnet agents, and our 18 mainnet agents plus Keel.

Data pulled Oct 8, 2026, from rpc.sentio.xyz, cross-checked on a second public RPC and against week one's data. Funding trails used Alchemy. Every cited transaction was re-read. "Exchange-style hot wallet" means 100,000+ sent transactions. We have no address labels. Counts are addresses, not people, and nothing here is a verdict on the campaign, any marketplace or any wallet.

Disclosure: ChainWard has an open BNB Chain Builder Grant application (submitted Oct 3) and sent BNB Chain a paid verification offer on Oct 5 (no reply as of Oct 7). It sells a $0.10 BSC hire check over x402 and lists a 0.25 USDT hire check on TermiX under its own ERC-8004 agent 365669, registered Oct 7. That agent is in this decode's registration count and had no hires (0 TermiX orders as of 04:31 UTC Oct 8). The board counts it in its registration total, never in its hirer or passing totals. In September, ChainWard offered TermiX the wallet-level version of its audit as a private, paid report. ChainWard also sells the September wallet list whose groups this decode uses, published the week-one decode and runs the free Set and Earn board. ChainWard has not entered Set and Earn as a participant.