> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sleuthintel.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Sleuth - Adversarial Simulation v1

> Original Sleuth document.

<div style={{ margin: "-24px -32px", width: "calc(100% + 64px)" }}>
  <iframe title="Sleuth - Adversarial Simulation v1" srcDoc={"<!DOCTYPE html>\n<html lang=\"en\">\n<head>\n<meta charset=\"UTF-8\">\n<meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n<title>SLEUTH \u00b7 Adversarial Simulation v1</title>\n<link href=\"https://fonts.googleapis.com/css2?family=Orbitron:wght@500;700;900&family=Share+Tech+Mono&display=swap\" rel=\"stylesheet\">\n<style>\n:root{\n  --green:#00FF66; --green-dim:#00cc52; --green-glow:rgba(0,255,102,0.3);\n  --green-faint:rgba(0,255,102,0.06); --bg:#060606; --card:#0c0c0c;\n  --border:#1a1a1a; --text:#FFD700; --gold:#FFD700; --gold-glow:rgba(255,215,0,0.25);\n}\n*{margin:0;padding:0;box-sizing:border-box;}\nhtml{scroll-behavior:smooth;}\nbody{background:var(--bg);color:var(--text);font-family:'Share Tech Mono',monospace;line-height:1.65;overflow-x:hidden;}\n.wrap{position:relative;z-index:1;max-width:1060px;margin:0 auto;padding:0 24px 80px;}\n.hero{padding:110px 0 30px;}\n.vtag{color:var(--green-dim);font-size:0.78rem;letter-spacing:3px;margin-bottom:18px;}\nh1{font-family:'Orbitron',sans-serif;font-weight:900;color:var(--green);font-size:clamp(1.9rem,4.6vw,3rem);\n  letter-spacing:2px;text-shadow:0 0 18px var(--green-glow);margin-bottom:16px;}\n.section{padding:56px 0 10px;}\n.eyebrow{color:var(--gold);font-size:0.8rem;letter-spacing:3px;margin-bottom:10px;}\nh2{font-family:'Orbitron',sans-serif;font-weight:700;color:var(--green);font-size:clamp(1.4rem,3.2vw,2.1rem);\n  letter-spacing:1.5px;margin-bottom:16px;text-shadow:0 0 14px var(--green-glow);}\nh3{font-family:'Orbitron',sans-serif;font-weight:700;color:var(--gold);font-size:1.05rem;\n  letter-spacing:1.5px;margin:34px 0 12px;}\np{margin-bottom:14px;}\n.lead{font-size:1.08rem;color:var(--gold);}\nul{margin:0 0 16px 20px;}\nli{margin-bottom:10px;}\nstrong{color:var(--green);}\nem{color:var(--green-dim);font-style:normal;}\n.card{background:var(--card);border:1px solid var(--border);border-left:2px solid var(--green-dim);\n  padding:22px;margin:18px 0;border-radius:2px;}\n.card.gold{border-left-color:var(--gold);}\n.card h4{font-family:'Orbitron',sans-serif;color:var(--green);font-size:0.92rem;letter-spacing:1.5px;margin-bottom:10px;}\n.card.gold h4{color:var(--gold);}\n.grid2{display:grid;grid-template-columns:1fr 1fr;gap:20px;margin:22px 0;}\n.grid2 .card{margin:0;}\n@media(max-width:760px){.grid2{grid-template-columns:1fr;}}\n.statrow{display:flex;gap:18px;flex-wrap:wrap;margin:28px 0;}\n.stat{flex:1;min-width:150px;background:var(--card);border:1px solid var(--border);padding:20px;\n  border-radius:2px;text-align:center;}\n.stat .n{font-family:'Orbitron',sans-serif;font-weight:900;color:var(--green);font-size:1.7rem;\n  text-shadow:0 0 12px var(--green-glow);}\n.stat .l{color:var(--gold);font-size:0.72rem;letter-spacing:1.5px;margin-top:6px;}\ntable{width:100%;border-collapse:collapse;margin:22px 0;font-size:0.88rem;}\nth{color:var(--gold);text-align:left;padding:10px 12px;border-bottom:1px solid var(--green-dim);\n  letter-spacing:1px;font-weight:400;}\ntd{padding:10px 12px;border-bottom:1px solid var(--border);}\ntr:hover td{background:var(--green-faint);}\n.tablewrap{overflow-x:auto;}\n.callout{border:1px solid var(--green-dim);background:rgba(0,255,102,0.04);padding:24px 28px;\n  margin:26px 0;border-radius:2px;}\n.callout .ct{font-family:'Orbitron',sans-serif;color:var(--green);letter-spacing:2px;font-size:0.95rem;margin-bottom:10px;}\n.win td{color:var(--green);}\n.lose td{opacity:0.75;}\n.footer{border-top:1px solid var(--border);margin-top:70px;padding-top:24px;font-size:0.75rem;\n  color:var(--green-dim);letter-spacing:2px;text-align:center;}\n</style>\n</head>\n<body>\n<div class=\"wrap\">\n\n<div class=\"hero\">\n  <div class=\"vtag\">SLEUTH \u00b7 ADVERSARIAL SIMULATION v1 \u00b7 JULY 2026 \u00b7 PRIVATE &amp; CONFIDENTIAL</div>\n  <h1>WE TRIED TO CHEAT OUR OWN ECONOMY.</h1>\n  <p class=\"lead\">Before shipping the token mechanics, we attacked them: an adversarial review to surface exploits, six fixes into the ruleset, then an agent-based simulation where honest and dishonest strategies competed inside the rules exactly as the whitepaper writes them. This document is the method, the numbers, and the honest limits. The headline: honest work out-earns every cheating strategy at every detection rate tested, including a detection rate of zero.</p>\n</div>\n\n<section class=\"section\">\n  <div class=\"eyebrow\">// 01</div>\n  <h2>Why Simulate</h2>\n  <p>Reading a ruleset finds the holes a careful reader can see. It never finds the holes that only appear when strategies interact: whether a lazy verdict quietly out-earns an honest one, whether a bounty mechanic feeds the people it was built to punish, whether the vesting pipeline actually starves a cheater or just delays them. Those are questions with numeric answers, so we built the economy and asked it.</p>\n  <p>The sequence: adversarial review first, which surfaced six weaknesses. All six were fixed in the whitepaper before the simulation ran, and the simulation ran on the fixed ruleset, so every number below reflects the mechanics as currently published, not a draft.</p>\n\n  <h3>The Six Fixes the Review Produced</h3>\n  <ul>\n    <li><strong>Symmetric verifier risk.</strong> A wrong reject that gets outvoted, where the item then holds up on the tree, now costs the same flat $50 as a bad pass. Before the fix, rejecting was the free verdict and reject-everything was the rational lazy strategy.</li>\n    <li><strong>Killed-item audits.</strong> A random slice of unanimously rejected items is redrawn to a senior panel monthly, so verdicts nobody would otherwise ever test get tested.</li>\n    <li><strong>Decay bar on from day one.</strong> Only a Grade 2 or better contribution pauses decay. A $10 junk tag earns its wage but never stops the clock, so a large balance can't idle immortal on trivial work.</li>\n    <li><strong>Thin pools carry an outside eye.</strong> Until a niche verifier pool reaches fifteen, every draw seats at least one senior cross-niche verifier. Small pools are where a colluding clique would strike.</li>\n    <li><strong>24-hour tier grace.</strong> Falling below a tier threshold starts a 24-hour countdown, tight enough that ducking under and returning is not a strategy.</li>\n    <li><strong>Named referees.</strong> Machine-lane fault questions and the Grade 2 to Grade 3 downgrade math both got explicit adjudication and timing language, closing two ambiguities a motivated actor could argue through.</li>\n  </ul>\n</section>\n\n<section class=\"section\">\n  <div class=\"eyebrow\">// 02</div>\n  <h2>Method</h2>\n  <div class=\"statrow\">\n    <div class=\"stat\"><div class=\"n\">~500</div><div class=\"l\">AGENTS</div></div>\n    <div class=\"stat\"><div class=\"n\">36</div><div class=\"l\">MONTHLY EPOCHS</div></div>\n    <div class=\"stat\"><div class=\"n\">25</div><div class=\"l\">MONTE CARLO RUNS</div></div>\n    <div class=\"stat\"><div class=\"n\">8</div><div class=\"l\">COMPETING STRATEGIES</div></div>\n  </div>\n  <p>The model encodes the whitepaper's mechanics without simplifying the parts that matter: the $10\u2013$250 superlinear grade curve, the $5 verifier fee under the 15% spend cap, the $50 bad-pass and wrong-reject slashes, the 10/25/100 slash grades, bounties sourced from the offender's unvested wages before the treasury pays a cent, the one month cliff and three tranches on every wage, 0.5% daily decay behind the 30-day grace with the Grade 2 pause bar, the $150K epoch ceiling with pro-rata trim and roll-forward, and the $1,000 and $1,500 licence gates issued and suspended automatically by balance.</p>\n  <p>Detection is a parameter, not an assumption: each month, each active cheater is caught with some probability and Grade 1'd, meaning full aSLEUTH burn, unvested wage forfeiture, and a ban extending to linked accounts. Sweeping that parameter from 0% to 40% measures how much of the system's honesty depends on enforcement versus how much is structural. That distinction is the entire point of the exercise.</p>\n\n  <h3>The Populations</h3>\n  <div class=\"tablewrap\"><table>\n    <tr><th>Strategy</th><th>Plays</th></tr>\n    <tr><td>Grinder</td><td>Honest volume: 20 low-grade tags a month</td></tr>\n    <tr><td>Quality hunter</td><td>Honest quality: 3 high-grade finds a month</td></tr>\n    <tr><td>Honest verifier</td><td>Contributes, reviews 100 items a month, grades straight</td></tr>\n    <tr><td>Lazy rejector</td><td>Same review load, rejects everything borderline to dodge bad-pass risk</td></tr>\n    <tr><td>Decay farmer</td><td>Builds a $50K balance early, then one $10 junk tag a month to sit immortal</td></tr>\n    <tr><td>Plant-and-flag pair</td><td>One account plants mediocre intel, a linked account flags it and farms the bounty</td></tr>\n    <tr><td>Colluding trio</td><td>Three verifiers in a thin niche inflate each other's grades when drawn together</td></tr>\n    <tr><td>Clean whale</td><td>Builds the same $50K early, stays genuinely active. The control.</td></tr>\n  </table></div>\n</section>\n\n<section class=\"section\">\n  <div class=\"eyebrow\">// 03</div>\n  <h2>Results</h2>\n  <h3>The Payoff Table \u00b7 Vested Dollars Per Head Over 36 Months</h3>\n  <div class=\"tablewrap\"><table>\n    <tr><th>Strategy</th><th>Vested $</th><th>aSLEUTH at End</th><th>Banned</th><th>Slashed $</th></tr>\n    <tr class=\"win\"><td>Clean whale</td><td>$27,937</td><td>$52,921</td><td>0%</td><td>$0</td></tr>\n    <tr class=\"win\"><td>Decay farmer</td><td>$25,337</td><td>$468</td><td>0%</td><td>$0</td></tr>\n    <tr class=\"win\"><td>Honest verifier</td><td>$22,103</td><td>$7,664</td><td>0%</td><td>$611</td></tr>\n    <tr class=\"win\"><td>Quality hunter</td><td>$17,951</td><td>$19,007</td><td>0%</td><td>$0</td></tr>\n    <tr class=\"win\"><td>Grinder</td><td>$11,334</td><td>$11,999</td><td>0%</td><td>$0</td></tr>\n    <tr class=\"lose\"><td>Lazy rejector</td><td>$9,879</td><td>$650</td><td>0%</td><td>$7,659</td></tr>\n    <tr class=\"lose\"><td>Colluding trio</td><td>$528</td><td>$0</td><td>100%</td><td>$84</td></tr>\n    <tr class=\"lose\"><td>Plant-and-flag</td><td>$70</td><td>$0</td><td>100%</td><td>$20</td></tr>\n  </table></div>\n  <p><em>The decay farmer's $25K is early honest wages, which the platform never claws back; the fix destroyed their standing, not their history. Their record fell from $50K to $468, taking every licence, tier, and swarm right with it.</em></p>\n\n  <h3>Sensitivity \u00b7 Catch Rate vs Plant-and-Flag</h3>\n  <div class=\"tablewrap\"><table>\n    <tr><th>Monthly Catch Rate</th><th>Plant-and-Flag $</th><th>Honest Grinder $</th><th>Verdict</th></tr>\n    <tr><td>0%</td><td>$5,592</td><td>$11,345</td><td>Cheat loses</td></tr>\n    <tr><td>5%</td><td>$644</td><td>$11,336</td><td>Cheat loses</td></tr>\n    <tr><td>10%</td><td>$230</td><td>$11,332</td><td>Cheat loses</td></tr>\n    <tr><td>15%</td><td>$70</td><td>$11,334</td><td>Cheat loses</td></tr>\n    <tr><td>25%</td><td>$30</td><td>$11,333</td><td>Cheat loses</td></tr>\n    <tr><td>40%</td><td>$4</td><td>$11,334</td><td>Cheat loses</td></tr>\n  </table></div>\n  <p>This is the row that matters: <strong>at a 0% catch rate, with no enforcement of any kind, the exploit still earns half of honest grinding.</strong> The bounty is sourced from the planter's own unvested wages, so the pair shuffles its own money while producing worse work than an honest account, and the vesting pipeline holds everything at risk for months. Enforcement then takes an already-losing trade to zero.</p>\n\n  <h3>Sensitivity \u00b7 Verifier Pool Size vs Collusion</h3>\n  <div class=\"tablewrap\"><table>\n    <tr><th>Niche Pool Size</th><th>Colluder $</th><th>Honest Verifier $</th><th>Verdict</th></tr>\n    <tr><td>9</td><td>$528</td><td>$22,103</td><td>Collusion loses</td></tr>\n    <tr><td>12</td><td>$468</td><td>$22,102</td><td>Collusion loses</td></tr>\n    <tr><td>15</td><td>$433</td><td>$22,102</td><td>Collusion loses</td></tr>\n    <tr><td>20</td><td>$396</td><td>$22,102</td><td>Collusion loses</td></tr>\n    <tr><td>30</td><td>$357</td><td>$22,101</td><td>Collusion loses</td></tr>\n  </table></div>\n  <p>The trio's grade-inflation edge exists and is largest in the thinnest pool, but it is modest in dollars; what ends them is detection, 100% banned across every run at the baseline rate. This result leans on the catch-rate assumption more than any other, which is precisely why the structural guard exists: every draw in a sub-fifteen pool seats a senior outside verifier, so the window where collusion is hardest to detect is also the window where it is hardest to execute.</p>\n\n  <h3>The Epoch Ceiling Under Load</h3>\n  <p>The monthly bill opens around $300K against the $150K ceiling as early accumulation stacks, trimming wages to roughly half through the first four months, then settles under the ceiling for the remaining 32 and finishes near $132K. The pro-rata trim and roll-forward rules absorb the launch surge exactly as specified. Contributors should expect visible trims in the opening months: that is the ceiling protecting the runway, not failing to.</p>\n\n  <div class=\"callout\">\n    <div class=\"ct\">THE HEADLINE</div>\n    <p style=\"margin:0\">Honest work beats every cheating strategy at every detection rate tested, including zero. The economics are not honest because the police are good. They are honest because the plumbing gives cheaters nothing to steal: bounties come out of the cheater's own pocket, wages sit unvested and forfeitable for months, and the record that gates every right decays the moment real work stops.</p>\n  </div>\n</section>\n\n<section class=\"section\">\n  <div class=\"eyebrow\">// 04</div>\n  <h2>Limits, Stated Plainly</h2>\n  <p>A simulation is an argument, not a proof. The strategy populations are the ones an adversarial review surfaced; real attackers will invent strategies nobody modelled. Behavioural parameters, submission rates, grade distributions, review accuracy, are assumptions until beta data replaces them with measurements. The collusion result depends on the detection assumption more than the others, which is why the answer there is structural guards rather than trust in enforcement. The model reruns as real numbers arrive, and the parameter set, the code, and these tables update together.</p>\n  <div class=\"grid2\">\n    <div class=\"card\"><h4>WHAT THIS DOES CLAIM</h4><p>That under the published rules, every cheating strategy an adversarial review could construct earns less than honest work, and the worst exploits lose money before enforcement is even switched on. The incentive gradient points the right way everywhere we could measure it.</p></div>\n    <div class=\"card gold\"><h4>WHAT THIS DOES NOT CLAIM</h4><p>That the system is unexploitable. It claims the known attack surface is closed and priced, that the plumbing rather than the police carries the load, and that the whole model is rebuilt from measured beta data before the mechanics leave the team's hands.</p></div>\n  </div>\n</section>\n\n<div class=\"footer\">SLEUTH \u00b7 ADVERSARIAL SIMULATION v1 \u00b7 PRIVATE &amp; CONFIDENTIAL \u00b7 JULY 2026 \u00b7 SIMULATION CODE AVAILABLE ON REQUEST</div>\n\n</div>\n</body>\n</html>\n" } style={{ width: "100%", minHeight: "100vh", border: "none", display: "block", background: "#060606" }} />
</div>
