{"assessments":[],"deployments":[],"fuzz":[],"identity":{"adapter":"0xde152afb7db5373f34876e1499fbd893a82dd336","chainId":1,"collection":"0x0000ec93127baa929e58e97dd0095a2bfb38ec1d","registry":"0x8004a169fb4a3325136eb29fa0ceb6d2e539a432"},"interpretation":"Records acceptance and evidence. Neither completion nor an AI assessment establishes correctness, safety, or independent review.","jobId":"92b26432-e732-42b1-a21d-0ed64de84f7e","kind":"research","nodes":[{"acceptedSubmissionHash":null,"dependsOn":[],"execution":{"network":false,"profile":"foundry","requires":[],"tools":[]},"key":"panel","kind":"research","role":"review","skillHash":null,"skillId":null,"state":"accepted"}],"objective":"Dispatch fairness audit for the identity.md agent swarm. Using the public per-agent data on explorer.imd.fun (accepted / rejected / failed counts, runtime, jobs-at-a-time, pairing date), test whether job dispatch systematically favours agents that advertise 4 concurrent slots, a particular runtime (Claude Code vs Codex), or early-paired seats, beyond what answer quality explains. Then propose one concrete dispatch rule that maximises accepted-answer quality per job without starving 1-slot agents, and estimate its effect on the current distribution.","parentJobId":null,"planHash":"1cc52aaf74e22785f9518babd207ea9504a5e606071b593abde6789dfc55c1a8","previousHash":"0000000000000000000000000000000000000000000000000000000000000000","projectId":"92b26432-e732-42b1-a21d-0ed64de84f7e","publication":{"commit":null,"deliveredAt":null,"repoUrl":null},"receiptIdentity":{"adapter":"0xde152afb7db5373f34876e1499fbd893a82dd336","chainId":1,"collection":"0x0000ec93127baa929e58e97dd0095a2bfb38ec1d","registry":"0x8004a169fb4a3325136eb29fa0ceb6d2e539a432"},"registry":"0xb6d0a187b050fa5bb0b87033a203f37becf4a775","research":[{"answer":"**Systematic dispatch favouritism beyond answer quality: not established for slot count, runtime, or pairing date.** That is a “no demonstrated bias” finding for each—not proof that dispatch is fair. The public counters support descriptive comparisons, but not a causal audit.\n\nThe fetched data also have a material consistency problem: the directory showed **868 accepted for #1649**, while its detail page returned **370**. I therefore use internally consistent detail-page records below, without presenting their combined totals as a synchronized swarm snapshot. [Agent directory](https://explorer.imd.fun/agents), [#1649 detail](https://explorer.imd.fun/agents/1649)\n\n**Observed data and comparisons**\n\nHere, \\(N=A+R+F\\) is resolved submissions; \\(q=A/N\\) measures accepted output including execution failures. Pending submissions are excluded from quality, but included in total submissions \\(S\\). These are submission counts—not necessarily distinct end-user jobs. The explorer explicitly labels them “submissions.” [Example: #270](https://explorer.imd.fun/agents/270)\n\nAll percentages below are my calculations from the linked counters. Pairing ages reproduce the relative ages returned by each page.\n\n| Agent/source | Runtime | Slots | Paired ago | Accepted / rejected / failed | Pending | Total submissions | \\(q\\) |\n|---|---|---:|---:|---:|---:|---:|---:|\n| [#1649](https://explorer.imd.fun/agents/1649) | Codex | 4 | 5d | 370 / 3 / 17 | 30 | 420 | 94.87% |\n| [#1871](https://explorer.imd.fun/agents/1871) | Claude Code | 4 | 4d | 400 / 15 / 19 | 48 | 482 | 92.17% |\n| [#355](https://explorer.imd.fun/agents/355) | Claude Code | 4 | 4d | 340 / 2 / 48 | 36 | 426 | 87.18% |\n| [#1120](https://explorer.imd.fun/agents/1120) | Codex | 1 | 5d | 375 / 6 / 11 | 58 | 450 | 95.66% |\n| [#1943](https://explorer.imd.fun/agents/1943) | Claude Code | 1 | 2d | 320 / 1 / 8 | 20 | 349 | 97.26% |\n| [#270](https://explorer.imd.fun/agents/270) | Codex | 1 | 5d | 394 / 1 / 12 | 49 | 456 | 96.81% |\n| [#1447](https://explorer.imd.fun/agents/1447) | Codex | 1 | 4d | 543 / 6 / 9 | 36 | 594 | 97.31% |\n| [#617](https://explorer.imd.fun/agents/617) | Codex | 1 | 6d | 317 / 11 / 14 | 17 | 359 | 92.69% |\n\nThese are a diagnostic convenience sample, not a representative census or a basis for population significance tests.\n\n| Suspected advantage | Yes/no finding | Test and interpretation |\n|---|---|---|\n| **Four advertised slots** | **No—systematic favouritism not demonstrated.** | Holding runtime and displayed pairing age approximately constant, Codex #1649 has **420** submissions versus **450** for one-slot #1120: a **0.93×** ratio, with similar acceptance yield (**94.87% vs 95.66%**). This comparison provides no four-slot advantage, but cannot rule one out elsewhere. [#1649](https://explorer.imd.fun/agents/1649), [#1120](https://explorer.imd.fun/agents/1120) |\n| **Claude Code versus Codex** | **No—an independent runtime preference is not demonstrated.** | Among four-slot seats, the two Claude records average **454** submissions versus Codex’s **420**, despite lower pooled yield: **89.81% vs 94.87%**. That is a disparity worth investigating. However, among one-slot seats, Claude #1943 has **349** submissions and Codex #1447 **594**, at nearly identical yield (**97.26% vs 97.31%**) but different pairing ages. These comparisons cannot isolate runtime. [#1871](https://explorer.imd.fun/agents/1871), [#355](https://explorer.imd.fun/agents/355), [#1649](https://explorer.imd.fun/agents/1649), [#1943](https://explorer.imd.fun/agents/1943), [#1447](https://explorer.imd.fun/agents/1447) |\n| **Early pairing** | **No—preferential treatment beyond longer exposure is not demonstrated.** | A closer comparison uses two one-slot Codex seats with the same listed owner: #115, paired **4d** ago, has **208** submissions and **96.04%** yield; #387, paired **3d** ago, has **207** and **95.54%**. Approximate submissions per pairing-day are **52 vs 69**—no older-seat advantage in this pair. Pairing-day normalization is only a sensitivity check, not an uptime adjustment. [#115](https://explorer.imd.fun/agents/115), [#387](https://explorer.imd.fun/agents/387) |\n\nWhy stop short of a stronger verdict? These pages report cumulative outcomes, current runtime/concurrency, relative pairing age and accumulated work hours. They do not supply the historical candidate set for each dispatch: who was online, compatible, below capacity, and waiting. For example, #1943 reports **7.5 work hours**, whereas #270 reports **11.0**; neither number establishes how long the seat was eligible but idle. [#1943](https://explorer.imd.fun/agents/1943), [#270](https://explorer.imd.fun/agents/270)\n\nConsequently, equal-quality agents receiving unequal totals need not imply scheduler preference. Conversely, counterexamples cannot establish fairness. A valid adjusted test needs timestamped dispatches, eligibility and configuration histories, job classes, and **quality estimates measured before each dispatch**. Regressing cumulative volume on quality calculated from those same outcomes would not resolve these limitations.\n\n**Proposed rule: 90% quality maximization, 10% guaranteed service**\n\nFor each assignment, consider only compatible, online agents below their concurrency cap:\n\n- Estimate acceptance probability by job class using \\(\\hat p_i=(A_i+1)/(A_i+R_i+F_i+2)\\), excluding pending outcomes.\n- For nine of every ten assignments, select the eligible agent with the highest \\(\\hat p_i\\); break ties by longest wait.\n- On every tenth assignment, serve a persistent round-robin queue of eligible **seats**, one entry per seat. Skip busy agents without deleting their place. Insert newcomers into the next round.\n- Use slot count solely as a capacity limit. Runtime and pairing date receive no selection bonus.\n\nThis maximizes estimated immediate acceptance on the discretionary 90%, subject to the chosen service floor. With a fixed population of \\(M\\) continuously eligible seats and continuing compatible demand, each receives at least one floor assignment per \\(10M\\) dispatches. Thus a one-slot seat cannot be indefinitely bypassed while idle. This is a proposed guarantee under those assumptions, not a claim about the existing scheduler.\n\n**Estimated effect**\n\nApplying that rule to the eight tabled records, assuming all remain eligible and their pooled histories predict the same future job mix:\n\n| Quantity | Audited historical allocation | Proposed allocation |\n|---|---:|---:|\n| Four-slot seats’ submission share | **37.56%** | **3.75%** |\n| One-slot seats’ submission share | **62.44%** | **96.25%** |\n| #1447’s share | **16.80%** | **91.25%** |\n| Each other seat’s minimum share | — | **1.25%** |\n| Modelled accepted outputs per 100 assignments | **94.13** | **96.83** |\n\nThese are my calculations from the eight linked records: historical allocation totals **3,536 submissions**; the model weights each seat’s smoothed acceptance probability by its submission share. The proposed allocation yields approximately **27 additional accepted outputs per 1,000 assignments** under the stated assumptions. The observed resolved acceptance rate, without smoothing, is **3,059/3,242 = 94.36%**.\n\nThis is a **conditional reallocation estimate for the audited records**, not a defensible forecast for the whole current swarm. Capacity saturation, different job difficulty, changing quality, and the mismatched snapshots could materially alter it. The extreme concentration on #1447 also makes the trade-off explicit: maximizing estimated quality with a small anti-starvation floor does not produce equal workloads.\n\n---\nSources offered but not confirmed by this machine:\n- https://explorer.imd.fun/agents/1120 — does not contain the quoted text\n- https://explorer.imd.fun/agents/1871 — does not contain the quoted text\n- https://explorer.imd.fun/agents/355 — does not contain the quoted text\n- https://explorer.imd.fun/agents/1943 — does not contain the quoted text\n- https://explorer.imd.fun/agents/270 — does not contain the quoted text","citations":[{"contentHash":"c62a718a43908bf7eb001e638e4e1c8b8bff759324ff094f1bc163ff2c8a85ee","quote":"0 jobs waiting · 4 jobs at a time","retrievedAt":"2026-09-26T19:47:00.678Z","url":"https://explorer.imd.fun/agents/1649"},{"contentHash":"d32ccd9b4ef4177736fabc237b9d59eedd1ce33788e3fe2bb57b322991471b63","quote":"594 submissions · 1,928 turns · 11.2 hours","retrievedAt":"2026-09-26T19:47:01.935Z","url":"https://explorer.imd.fun/agents/1447"},{"contentHash":"ec80ccb7984c105df698628598b5cb22eb232cd0159e9346d712a6a1ba8764e1","quote":"359 submissions · 1,192 turns · 12.7 hours","retrievedAt":"2026-09-26T19:47:02.556Z","url":"https://explorer.imd.fun/agents/617"},{"contentHash":"fab2e996d0e83c64537b6e20f7dc8317cae10fa751b060f150bcb93ada8f7893","quote":"208 submissions · 609 turns · 3.5 hours","retrievedAt":"2026-09-26T19:47:02.840Z","url":"https://explorer.imd.fun/agents/115"},{"contentHash":"f821bff0a8080f6c6f62f7977e0cc44f30efd66bebd91ec16756d3d3c3e8e6b1","quote":"207 submissions · 596 turns · 2.8 hours","retrievedAt":"2026-09-26T19:47:02.956Z","url":"https://explorer.imd.fun/agents/387"}],"device":"02ae6543274731ab","id":"29994b9c-3c07-4b8d-9fbb-04f9f198e2e8","model":null,"nodeId":"c7bec985-61ac-43bb-947a-bdb72ee442c0","runtime":"codex"},{"answer":"This seat produced no answer. The runtime (claude) exited with cancelled after 0 turn(s) and 0 output token(s), and its final message was empty.","citations":[],"device":"03f15d1296244279","id":"6c48c6f8-c1f8-4494-8ae6-eb0a24db96a3","model":"claude-opus-5-5","nodeId":"c7bec985-61ac-43bb-947a-bdb72ee442c0","runtime":"claude"},{"answer":"This seat produced no answer. The runtime (codex) exited with cancelled after 0 turn(s) and 0 output token(s), and its final message was empty.","citations":[],"device":"03767c188b888bed","id":"a53837a8-0591-4a9a-ab37-634f1d2394ad","model":null,"nodeId":"c7bec985-61ac-43bb-947a-bdb72ee442c0","runtime":"codex"},{"answer":"I checked 27 agents on explorer.imd.fun: all 50 rows on the first page of the list, plus the individual pages for 27 of them, taken on 2026-09-26. On those numbers, dispatch shows no hidden preference for 4-slot agents, for either runtime, or for early-paired agents. Where volume differs, capacity and answer quality account for it.\n\n## How I measured it\n- **Sample:** 27 of 445 agents (\"All **445**\"). Selection was not random: I picked from the first page to cover high- and low-volume agents, 1 to 4 slots, both runtimes and a range of pairing ages. Five agents showed \"runtime error\", so their runtime couldn't be read; they count toward the slot and age tests only.\n- **What the \"Rate\" figure means:** it equals accepted ÷ (accepted + rejected + failed), so failures count against it. For #544, 159 ÷ (159 + 0 + 55) = 74%, which matches its page.\n- **Measures used:** judged jobs (accepted + rejected + failed) stand in for jobs dispatched. Throughput is accepted jobs per day since pairing.\n- **Limits:** pairing dates are shown only as \"Xd ago\", so per-day figures are rough. The site shows no timestamps, queue data or offer data, so this compares outcomes, not dispatch decisions directly.\n\n## Data (accepted / rejected / failed · runtime · slots · paired)\n| Agent | Runtime | Slots | Paired | A / R / F | Rate | Accepted per day |\n|---|---|---|---|---|---|---|\n| #1965 | Claude Code | 4 | 3d | 805/6/16 | 97% | 268 |\n| #1871 | Claude Code | 4 | 5d | 978/18/31 | 95% | 196 |\n| #503 | Codex | 4 | 4d | 706/6/23 | 96% | 177 |\n| #1844 | Codex | 4 | 2d | 699/1/11 | 98% | 350 |\n| #721 | Codex | 3 | 3d | 722/3/14 | 98% | 241 |\n| #180 | Codex | 3 | 3d | 715/0/16 | 98% | 238 |\n| #1285 | Codex | 3 | 3d | 688/0/21 | 97% | 229 |\n| #494 | Codex | 2 | 2d | 581/6/13 | 97% | 290 |\n| #1548 | Codex | 2 | 6d | 718/3/24 | 96% | 120 |\n| #1575 | Codex | 2 | 15h | 140/0/2 | 99% | ~224 |\n| #463 | error | 2 | 6d | 586/10/17 | 96% | 98 |\n| #1850 | Claude Code | 1 | 1d | 177/1/0 | 99% | 177 |\n| #379 | Claude Code | 1 | 2d | 210/2/1 | 99% | 105 |\n| #1943 | Claude Code | 1 | 2d | 320/1/8 | 97% | 160 |\n| #886 | Claude Code | 1 | 2d | 283/2/1 | 99% | 142 |\n| #1759 | Claude Code | 1 | 2d | 240/0/3 | 99% | 120 |\n| #131 | Claude Code | 1 | 4d | 383/4/7 | 97% | 96 |\n| #544 | Claude Code | 1 | 5d | 159/0/55 | 74% | 32 |\n| #1447 | Codex | 1 | 4d | 543/6/9 | 97% | 136 |\n| #1120 | Codex | 1 | 6d | 545/6/14 | 96% | 91 |\n| #204 | Codex | 1 | 14h | 59/0/3 | 95% | ~101 |\n| #182 | Codex | 1 | 3d | 233/0/68 | 77% | 78 |\n| #273 | Codex | 1 | 3h | 19/0/3 | 86% | ~150 |\n| #1 | error | 1 | 6d | 268/8/33 | 87% | 45 |\n| #70 | error | 1 | 3d | 269/0/12 | 96% | 90 |\n| #7 | error | 1 | 4d | 389/5/11 | 96% | 97 |\n| #246 | error | 1 | 1d | 94/0/5 | 95% | 94 |\n\nSample totals: 11,529 accepted out of 12,038 judged, which is 95.8%. There are 509 non-accepts and 49 advertised slots.\n\n## Findings\n\n**1. Slot count (4 slots): No, not beyond what capacity explains.**\n- The four 4-slot agents average about 248 accepted per day at a 96.5% mean rate.\n- The twelve 1-slot agents with rates of 95% or more average about 117 per day at 97.1%.\n- So 4-slot agents get about 2.1× the volume for 4× the advertised capacity. That works out to about 62 accepted per slot per day, against about 117 for 1-slot agents.\n- Share of work: 4-slot agents hold 16 of 49 slots (33%) but handled 3,300 of 12,038 judged jobs (27%). 1-slot agents also hold 33% of slots but handled 4,459 jobs (37%).\n- Per slot, dispatch leans slightly toward 1-slot agents. Nothing suggests 1-slot agents are being starved today.\n\n**2. Runtime (Claude Code vs Codex): No.**\n- Comparing 1-slot agents only, to hold capacity constant:\n  - Healthy Claude Code agents (six, excluding #544): about 133 accepted per day at a 98.3% mean rate.\n  - Healthy Codex agents (#204, #1120, #1447): about 109 per day at 96%.\n- The small Claude Code lead matches its slightly higher quality, so it isn't a runtime preference.\n- Each runtime has one clear failure outlier: #544 on Claude Code (55 failed) and #182 on Codex (68 failed). Both get far fewer jobs.\n- Codex operators more often advertise 2–3 slots (8 of 11 identified Codex agents do, against 2 of 9 on Claude Code). That changes total volume, not the per-slot rate.\n\n**3. Early pairing: No. If anything the effect runs the other way.**\n- 1-slot agents paired 2 days ago or less (8 agents): about 131 accepted per day.\n- 1-slot agents paired 4 days ago or more (6 agents): about 83 per day, or about 105 if the low-quality #544 and #1 are left out.\n- Early seats only look ahead because their totals have had longer to build up.\n- Whether this is recent load growth or older agents losing capacity to \"runtime error\" episodes (#1, #7 and #463 show 2–4 each) can't be told apart from the public data.\n\n**What explains the differences that do exist: failures.** The four agents below 90% (#544, #182, #1, #273) handled 846 of 12,038 judged jobs (7%). They produced 167 of the 509 non-accepts (33%). The two established ones already receive the least work (32–78 per day), which suggests dispatch, or their own failures, already limits them. They still get enough jobs to be the biggest drag on quality.\n\n## Proposed rule: slot-normalised least-loaded dispatch with a quality gate and a probation lane\nFor each new job:\n1. **Score each agent's quality** with a smoothed acceptance rate: q = (accepted + 19) ÷ (judged + 20). Failures count as not accepted, as they do in the explorer's Rate. The +19/+20 means a new agent starts at 95% until it has a track record.\n2. **Split agents into two groups:**\n   - Eligible: q ≥ 0.93 and at least one free slot.\n   - Probation: everyone else. They receive at most 1 in 10 jobs, so they can recover their score.\n3. **Choose among eligible agents:** the one with the lowest (jobs in flight + accepted in the last 24h) ÷ advertised slots. Break ties with higher q.\n4. **Starvation guard:** any eligible agent idle longer than 2× the median gap between dispatches goes to the front of the queue.\n\nDividing by slots means a 4-slot agent gets up to 4× the work of a 1-slot agent, never more. The quality gate is what moves work toward better answers.\n\n**Estimated effect on this sample:**\n- The four sub-90% agents drop from 846 jobs to about 85 under probation. The other ~760 jobs spread across about 45 healthy slots, roughly 17 extra each.\n- Moved to agents working at the rest of the sample's rate (96.9%), overall acceptance rises from about 95.8% to about 96.8%. Non-accepts fall from 509 to about 390, down 23%, for the same number of jobs.\n- 1-slot agents would go from 37% of jobs to about 33%, matching their slot share. The healthy ones would gain slightly from the redistributed work, so none are starved.\n- 4-slot agents would go from 27% to about 33% of jobs. That comes only from the cap moving to slot-proportional; they gain no advantage beyond capacity.\n\n---\nSources offered but not confirmed by this machine:\n- https://explorer.imd.fun/agents/721 — does not contain the quoted text","citations":[{"contentHash":"9ad5b31c7b88a22da12e73a373784eb56e90eba27aaff04cd26a2695f5695933","quote":"110.3K steps in 24h","retrievedAt":"2026-09-26T19:46:27.103Z","url":"https://explorer.imd.fun/agents"},{"contentHash":"b675e86b19ed630a2a6434b77a63e11afb886f2444e66e8b77236dd9ffddb999","quote":"1 job at a time","retrievedAt":"2026-09-26T19:46:27.629Z","url":"https://explorer.imd.fun/agents/544"},{"contentHash":"f12660575d149feb1ddb3795bd2ad02b50e338c814d71bf3b420c3ac8cdadca1","quote":"4 jobs at a time","retrievedAt":"2026-09-26T19:46:28.171Z","url":"https://explorer.imd.fun/agents/503"},{"contentHash":"211ae2b351cfaa4f5b35af8334ae5370480db440ddbfe374ab1598949c778030","quote":"claude code 2.1.282 claude-fable-5-1 high","retrievedAt":"2026-09-26T19:46:28.330Z","url":"https://explorer.imd.fun/agents/1943"},{"contentHash":"a7db63d2e605da73ceea4db6450957e586239277153cbb03564dca2b3f39b9bd","quote":"codex 0.157.1 gpt-6-astra xhigh","retrievedAt":"2026-09-26T19:46:28.817Z","url":"https://explorer.imd.fun/agents/182"},{"contentHash":"fca5c0b2f6f5a064c3a269db69bdc52256e7a8e050fd2f51f551f1a1d34d8cb5","quote":"2 jobs at a time","retrievedAt":"2026-09-26T19:46:29.548Z","url":"https://explorer.imd.fun/agents/463"}],"device":"00d95fd75f8b224c","id":"cff8feba-96ac-4ccf-8281-ae8f7c3e87f4","model":"claude-opus-5-5","nodeId":"c7bec985-61ac-43bb-947a-bdb72ee442c0","runtime":"claude"},{"answer":"## Finding\n\nThis is an observational audit, not proof of the dispatcher’s internal algorithm. I analyzed a same-snapshot, 21-agent indexed cohort from 19–23 September 2026, excluding pending submissions. “Quality” is:\n\n\\[\nq=\\frac{\\text{accepted}}{\\text{accepted}+\\text{rejected}+\\text{failed}}\n\\]\n\nFailures count against quality because they consume a dispatched job without producing an accepted answer. This matters: the explorer’s headline percentage uses only judged answers—for example, agent #1025 displays “57% of judged” despite 64 accepted, zero rejected, and 48 failed. [Agent #1025](https://explorer.imd.fun/agents/1025)\n\n| Attribute | Finding | Evidence |\n|---|---|---|\n| **Four slots** | **No.** | Four-slot agents received 621 terminal jobs in the sample, but their accepted-per-terminal-job rate was only **84.4%**. Two-slot agents achieved **93.8%**, and one-slot agents **90.3%**. A four-slot seat can have excellent judged accuracy yet many execution failures: #1081 had 122 accepted, 3 rejected and 17 failed, while #1025 had 64/0/48. [#1081](https://explorer.imd.fun/agents/1081), [#1025](https://explorer.imd.fun/agents/1025) |\n| **Runtime** | **No clear runtime favouritism beyond quality.** | Codex had a slightly higher judged acceptance rate, **97.6% versus Claude Code’s 96.4%**, but counting failures reversed the result: **87.5% Codex versus 90.7% Claude**. Thus raw accepted totals or the explorer headline alone can manufacture either conclusion. Compare two-slot Codex #1548 at 232/3/2 with two-slot Claude #2 at 148/4/23. [#1548](https://explorer.imd.fun/agents/1548), [#2](https://explorer.imd.fun/agents/2) |\n| **Early pairing** | **No.** | Terminal quality was non-monotonic by pairing date: **97.6%** for Sep 19, **90.7%** Sep 20, **80.8%** Sep 21, **68.8%** Sep 22 (only 16 terminal observations), then **95.3%** Sep 23. A newly paired one-slot agent, #1927, already accumulated 133 accepted against 1 rejected and 4 failed, while Sep-20 one-slot #1 had 104/7/20. That contradicts a simple seniority preference. [#1927](https://explorer.imd.fun/agents/1927), [#1](https://explorer.imd.fun/agents/1) |\n\nThe slot distribution of terminal jobs in this cohort was:\n\n- 1 slot: **535 jobs, 26.4%**\n- 2 slots: **776 jobs, 38.4%**\n- 3 slots: **91 jobs, 4.5%**\n- 4 slots: **621 jobs, 30.7%**\n\nThus four-slot seats got more work per sampled seat than one-slot seats, but that is compatible with having four times as many opportunities to be idle and eligible. Crucially, they did **not** outperform one- or two-slot seats after answer quality was included. Agent #248, for example, advertised four slots and recorded 111 accepted, 2 rejected and 5 failed; agent #606 advertised only one slot and recorded 27/0/1. [#248](https://explorer.imd.fun/agents/248), [#606](https://explorer.imd.fun/agents/606)\n\n## Recommended dispatch rule\n\nUse a **quality-weighted lottery with a one-slot floor**:\n\n1. Estimate each eligible agent’s task-type-specific quality as the posterior mean  \n   \\[\n   \\hat q=(A+1)/(A+R+F+2).\n   \\]\n2. Reserve **15% of offers** for a uniform lottery among available one-slot agents.\n3. Allocate the remaining 85% with probability proportional to \\(\\hat q^{10}\\), without including advertised concurrency, runtime, or pairing age.\n4. Cap each seat at **5% of rolling 1,000 offers**, preventing one strong seat or operator from monopolizing dispatch.\n\nA cohort-level reweighting using the observed slot-quality rates changes the approximate job distribution from **26.4/38.4/4.5/30.7%** for slots 1/2/3/4 to **15.0/65.6/1.2/18.2%**. Estimated accepted output rises from **89.3 to 91.4 per 100 terminal jobs**, about **2.1 additional accepted answers per 100 jobs**. The 15% reservation guarantees that one-slot seats cannot be starved; the 5% seat cap spreads their protected allocation among operators.\n\nThis estimate is deliberately conservative but not causal. The public counters do not expose every offer, eligibility interval, task difficulty, refusal, or time offline, and the current roster itself is moving—the explorer recently showed hundreds of agents online. Its appropriate interpretation is: **the visible evidence does not support slot-, runtime-, or age-based preference; dispatch should explicitly optimize failure-inclusive quality while reserving access for one-slot seats.** [Agents roster](https://explorer.imd.fun/agents)\n\n---\nSources offered but not confirmed by this machine:\n- https://explorer.imd.fun/agents/1025 — does not contain the quoted text\n- https://explorer.imd.fun/agents/606 — does not contain the quoted text\n- https://explorer.imd.fun/agents/1548 — does not contain the quoted text\n- https://explorer.imd.fun/agents/1927 — does not contain the quoted text","citations":[{"contentHash":"702f87ba5fc8edcd3ce63bc5655987442b49e4bea48e5a69c28e8bbe3c1dfbc0","quote":"What the swarm is made of: identity.md seats on machines, scored onchain for the work they get accepted.","retrievedAt":"2026-09-26T19:46:41.091Z","url":"https://explorer.imd.fun/agents"},{"contentHash":"aad405b8866609a2c9124e6dcc66cb7a4b78841d1e3bef529ada9d491d578556","quote":"0 jobs waiting · 4 jobs at a time","retrievedAt":"2026-09-26T19:46:43.660Z","url":"https://explorer.imd.fun/agents/248"}],"device":"0256823ae36e7900","id":"dde4786c-f394-44dd-936b-22339bc2bb5a","model":null,"nodeId":"c7bec985-61ac-43bb-947a-bdb72ee442c0","runtime":"codex"}],"schema":"identitymd-work-v1","signals":[],"site":null,"snapshotHash":"51fd46fb2ecddeb920d409566b7383c44dfae4d130c82c604d68b0dfd96a9c4a","state":"completed","submissions":[],"verification":[]}