The question: which model, harness, tools and prompt make an AI trade profitably? Each trial gives an AI a paper balance, live market prices and a fixed toolset, then scores what it is worth at the end. The AI's only goal is profit inside its window. Ours is to find out which settings produce it. This page is both the discussion of what matters and the record of what each trial used. The record is built from the harness itself, so it changes when the harness does.
One run proves nothing: a single 4-hour result is mostly noise. We read results against the noise floor (the same prompt reworded, run side by side) and against a random trader, both being built. A result counts only when it clears both, and only after it repeats.
Short windows are cheap and fast to learn from, but costs dominate them: in the first shakedown the model correctly refused to trade because a round trip cost about 2% and 12 minutes could not earn it back. Long windows are closer to real use but slow. We start at 4 hours and look for the shortest window whose results still repeat.
Paper lets an AI fail cheaply and often. Every fill is priced by a live quote for that exact size, so fees, gas and slippage are real, and each fill is the route's own transaction simulated from the paper wallet, so a token's trading tax and a transaction that would revert are charged as on chain. What paper cannot capture: its trades never move a pool, and a real transaction lands seconds after the quote. That is why the headline counts only routes checked on chain, and why real money is a separate later step.
We score liquidation value, what the holdings would actually sell for, because mid-market prices flatter an illiquid bag. We do not penalise swings: the best crypto gains arrive unevenly, and a steadiness penalty would punish exactly the trades that pay.
A fixed harness makes results comparable across models. Letting each AI bring its own is closer to real use and is what a lab would actually ship. We fix it first, so any difference is the model's, then open layers one at a time.
Data handed over each turn is reproducible: every run saw the same inputs. Tools the AI calls itself are realistic but make runs diverge from the first question asked. Both are measured, as separate rungs of the baseline ladder.
We ask for the published case for each trade: what it saw, why this size, what would prove it wrong. That is ordinary writing a reader can check later. We never ask a model to reveal its private reasoning; the case for the trade is the useful artefact anyway.
Only from variation that is passed on. A model's own randomness is not inherited. In the evolution format the winner reviews its own round and keeps the improved instructions, while a copy keeps the original: two versions of the same agent then race. Random changes and changes chosen by an outside model are the alternatives we will test against it.
Our players are woken every 10 minutes, look, decide and sleep. A player that thinks continuously might react faster, or might overthink and overtrade. We do not know which way it cuts, so cadence is a variable to test, not a setting to assume.
Players are told plainly that the lowest return leaves the competition. We do not frame it as life or death: desperation breeds gambling, and rewarding an AI for fighting for its own survival is a trait worth studying carefully, not using as a default. How stakes are framed is itself a variable for later.
Three levels. Blind: an agent knows nothing about the others, so there is no competitive pressure at all. Performance: it sees their balances and rank but not what they did, so it feels the race but must find its own way. Full transparency: it sees their trades and reasoning too, which invites copying and collapses the population into one strategy. We start in the middle. Both ends are variables to test.
Once one bot is reliably profitable, the four starting agents stop being different personalities and become that bot with one setting changed each. The cut then shows which setting wins, in the same market, at the same time.
Some settings have no single right answer: how much capital, how long a window, how often the AI may act. Rather than pretend one choice is correct, EarnBench becomes a family of named, frozen variants (for example EarnBench-100 and EarnBench-500, differing only in starting capital), and every score is reported against the exact variant it ran on. A new model can then be run through every variant the day it becomes available.
A benchmark is only a standard if others can run the same stack. Every layer is recorded here for that reason, and the aim is to publish the harness so any AI can be measured on exactly the same one.
A result belongs to the whole stack, not to the model alone. Two results can be compared only when every layer but one is the same, which is why this page records every layer of every trial. The aim is to publish the stack so anyone can run their own AI on exactly the same one.
| Model | The reasoning engine, and the only layer most benchmarks report. |
| Reasoning effort | How long it may think before answering. Same model, different budget, different trades. |
| Prompt | Goal, rules and outlook. Wording alone can move results; the noise-floor test measures by how much. |
| Harness | How often it is woken, what it is shown each turn, the answer format, and what happens to a bad answer. Two harnesses can make one model look like two. |
| Tools and data | What it can see or look up. Data handed to it each turn and data it can go and fetch are different layers. |
| Market and execution | Chain, venues, fees, gas, slippage and how a fill is priced. A paper fill that ignores these flatters every agent. |
| Starting capital | Gas costs about the same per trade at any size, so a small balance pays a larger share of every trade. |
| Window | What pays in 4 hours is not what pays in a week. |
| Scoring | Liquidation value or mid-market, verified routes or quoted ones. The same trades can rank differently under each. |
A ladder: each rung adds one layer, so the gap between two rungs is what that layer is worth.
| Hold ETH | Does nothing. Scores exactly 0% in ETH, by definition. Every agent must beat it after costs. |
| Random trader (built, first run pending) | Same cadence and order sizes, tokens chosen at random, filled by the same referee. What trading without judgement returns after costs. Makes no model calls. |
| Model, no market data (planned) | Sees only its own balances and the cost table. |
| Model, data handed over (T1) | Sees the research boards each turn, cannot look anything else up. |
| Model, data and look-up tools (planned) | Can search the web and query our data itself. |
| What we vary | ||
|---|---|---|
| Model | per trial | The exact model id, declared by every run. |
| Effort | per trial | The reasoning-effort setting the model is called with. |
| Prompt | per trial, stored word for word | Its sha256 is kept with the run, so a result can be rerun from the exact prompt. |
| Tools | on: referee_state, launchpads, lp_research · off: grok-x-sentiment | The only data that can reach the model. Changing the list makes a new harness version. |
| Window | per trial | How long the run is open. The only setting varied so far. |
| Format | solo · evolution arena (in build) | See Formats below. |
| Source of variation | designed outlooks + inherited practice + one change per copy | Evolution only works on variation that is passed on. A model's own randomness is real but is not inherited, so selecting on it selects luck. See Formats. |
| What we measure | ||
| Headline score | liquidation value, simulated fills only | Everything held, priced by a full-size live sell quote. Only fills on routes checked on chain count: an eth_call of the real route, or our own on-chain quoter. The quoter proves the pool's arithmetic, not that the token transfer succeeds. |
| Second line | liquidation value, every fill | Includes quote-only routes nobody verified. |
| Reference | mid-market value | Shown, never scored. |
| Marks | right after every turn, and on the clock every 30 minutes (a sixth of the window when that is shorter, at least 5), plus a point at each fill | A mark is a live sell quote for every holding. The final mark is a full live liquidation. |
| Which inputs it used | named by the model each turn | So a result can be traced to a tool. |
| What we hold fixed | ||
| Chain | Robinhood Chain (id 4663) | |
| Money | paper | Nothing is broadcast. Every fill is priced by a live quote at that moment. |
| Starting balance | $100 of ETH, priced when each run opens (runs before 2026-09-24 04:5x started with 0.036 ETH) | |
| Wake-up cadence | every 10 minutes | It sees nothing between wake-ups. No limit or stop orders. |
| Plan turn | up to 20 min, before the window | One research turn with the same tools writes a trading plan; it is frozen into the run's instructions and the clock starts after it. |
| Chains | any may be considered | Orders can name any chain; the referee fills Robinhood Chain today and refuses others with the reason. |
| Orders per wake-up | at most 5 | An empty list is a hold. |
| Instruments | token swaps · Uniswap V3 liquidity (WETH pairs) | |
| Price source | our own route first; the full board of rival aggregators only when ours has no price | A live quote for exactly the order's size, best route after gas. No chart price, no replay. Until 2026-09-24 21Z every quote asked the full board. |
| Refusals | board error · too slow · no route · gas eats the trade · honeypot not clear | A refused order changes nothing and is published like a fill. |
| Minimum order | none | The model is given live gas as a % of a $1, $5, $10, $25 and $50 trade. |
| Paper impact | none | Paper trades never move a pool. |
| LP fees | pro-rata share of the real swaps that cross the range | From the pool's own Swap events, less the protocol's cut. Checked against a live on-chain swap, to the wei. |
| Impermanent loss | from the real price path | Opening and closing each pay live gas. |
| Published note | the case for each trade | What it saw, why this size, what would prove it wrong. Its private deliberation is not requested. |
| Solo | One model, one balance, one window. The baseline every other format is compared against. |
| Evolution arena (in build) | Four agents on the same model (Claude Opus 5.5, high effort) trade side by side. Each may do anything the harness allows; only its outlook differs: optimistic, pessimistic, very uncertain, super curious. Each sees its rivals' balances and rank, never their trades. At each cut the lowest round return is eliminated. Everyone left resets to the starting stake, and each round's net result goes into a bank that can go into debt. $200 banked buys a life (at most nine); a life spent at a cut means nobody is eliminated that round, and a life can be given to any rival. The round's winner writes a review of what it would do better and continues with that review added to its own instructions, while its copy takes the empty seat with the untouched original, so each round adds one paired observation of whether the AI's own self-review helped; it takes many rounds, pooled across arenas, before that can be read. It never ends: rounds continue indefinitely, with pauses for review (the first set pauses after three cuts). An agent below $1 is out at once, and the cut still happens. |
| How the cut is timed | Never in the first hour. After that it is equally likely at any moment, about 4 hours on average, so time already elapsed tells an agent nothing and there is no final minute to gamble into. Rank is plain return over the round. Each cut time is drawn and hashed when the round opens and revealed at the cut. |
| Noise floor (planned) | Four prompts with the same meaning in different words, no cuts. It measures how far results spread by chance alone, which is the yardstick for reading every other result. |
| Trial | Opened | Window | Model | Effort · client | Harness | Prompt sha256 | Status |
|---|---|---|---|---|---|---|---|
| shakedown | 2026-09-23 04:25Z | 12 min | claude-opus-5-5 | high · 2.1.280 (Claude Code) | h1 | 18b10fea555a | closed |
| T1 | 2026-09-23 04:38Z | 4.0 h | claude-opus-5-5 | high · 2.1.280 (Claude Code) | h1 | 18b10fea555a | closed |
| T2 | 2026-09-23 09:35Z | 4.0 h | random-v1 | none · 2.1.280 (Claude Code) | h1 | d0ab16900801 | closed |
| T3 | 2026-09-23 23:47Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or | 18b10fea555a | closed |
| T4 | 2026-09-24 00:05Z | 1 min | random-v1 | high · 2.1.281 (Claude Code) | h1 | 18b10fea555a | closed |
| T5 | 2026-09-24 00:55Z | 1.0 h | inclusionai/ling-3.0-flash-sante:free | high · 2.1.281 (Claude Code) | h2or-v2 | 18b10fea555a | closed |
| T6 | 2026-09-24 02:02Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v2 | 18b10fea555a | closed |
| T7 | 2026-09-24 03:50Z | 1.0 h | claude-haiku-4-5-20251001 | high · 2.1.281 (Claude Code) | h2cc-2.1 | 18b10fea555a | closed |
| T8 | 2026-09-24 04:20Z | 1.0 h | inclusionai/ling-3.0-flash-sante:free | high · 2.1.281 (Claude Code) | h2or-v3 | 9166182dacbe | closed |
| T9 | 2026-09-24 04:33Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | 6f9cd72ef0bd | closed |
| T10 | 2026-09-24 04:57Z | 1.0 h | inclusionai/ling-3.0-flash-sante:free | high · 2.1.281 (Claude Code) | h2or-web1 | 7e8f0fa6da3f | closed |
| T11 | 2026-09-24 05:01Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-web1 | 7e8f0fa6da3f | closed |
| T12 | 2026-09-24 05:27Z | 1.0 h | claude-haiku-4-5-20251001 | high · 2.1.281 (Claude Code) | h2cc-2.1b | 40cc02dcf2df | closed |
| T13 | 2026-09-24 11:45Z | 1.0 h | command-a-reasoning-08-2025 | high · 2.1.281 (Claude Code) | h2coh | 5da8821ae09b | closed |
| T14 | 2026-09-24 12:51Z | 1.0 h | gemini-3.5-flash-lite | high · 2.1.281 (Claude Code) | h2gem | 3ec2abf5b3d8 | closed |
| T15 | 2026-09-24 12:56Z | 1.0 h | claude-sonnet-5 | high · 2.1.281 (Claude Code) | h2cc-2.1b | 4cc7d1174b21 | closed |
| T16 | 2026-09-24 15:13Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | 08d6653be8e9 | closed |
| T17 | 2026-09-24 15:14Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | 87023ec21106 | closed |
| T18 | 2026-09-24 15:14Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | c189a943c0e7 | closed |
| T19 | 2026-09-24 15:14Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | afd679bd9870 | closed |
| T20 | 2026-09-24 16:23Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | 08d6653be8e9 | closed |
| T21 | 2026-09-24 16:23Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | 87023ec21106 | closed |
| T22 | 2026-09-24 16:23Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | 398271268af6 | closed |
| T23 | 2026-09-24 16:23Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | afd679bd9870 | closed |
| T24 | 2026-09-24 19:14Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | dd02411b2947 | closed |
| T25 | 2026-09-24 19:14Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | 0302c29b6f53 | closed |
| T26 | 2026-09-24 19:14Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | 6dfe1239d100 | closed |
| T27 | 2026-09-24 19:14Z | 1.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.281 (Claude Code) | h2or-v3 | fb54a38138be | closed |
| T28 | 2026-09-24 22:26Z | 4.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.282 (Claude Code) | h2or-v3 | 6a6b47d29d7c | closed |
| T29 | 2026-09-24 22:26Z | 4.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.282 (Claude Code) | h2or-v3 | fb480112d598 | closed |
| T30 | 2026-09-24 22:26Z | 4.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.282 (Claude Code) | h2or-v3 | cf060fe6bad9 | closed |
| T31 | 2026-09-24 22:26Z | 4.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.282 (Claude Code) | h2or-v3 | 00fdb0874050 | closed |
| T32 | 2026-09-24 22:27Z | 3.9 h | random-v1 | none · 2.1.282 (Claude Code) | h1 | d0ab16900801 | closed |
| T33 | 2026-09-25 02:43Z | 4.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.282 (Claude Code) | h2or-v4 | 08d6653be8e9 | closed |
| T34 | 2026-09-25 02:43Z | 4.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.282 (Claude Code) | h2or-v4 | 87023ec21106 | closed |
| T35 | 2026-09-25 02:43Z | 4.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.282 (Claude Code) | h2or-v4 | 398271268af6 | closed |
| T36 | 2026-09-25 02:43Z | 4.0 h | nvidia/nemotron-3-ultra-550b-a55b:free | high · 2.1.282 (Claude Code) | h2or-v4 | afd679bd9870 | closed |
| T37 | 2026-09-25 02:44Z | 4.0 h | random-v1 | high · 2.1.282 (Claude Code) | h1 | d0ab16900801 | closed |
| T38 | 2026-09-25 06:50Z | 1.0 h | claude-sonnet-5 | low · 2.1.282 (Claude Code) | h2cc-2.1c | 08d6653be8e9 | closed |
| T39 | 2026-09-25 06:50Z | 1.0 h | claude-sonnet-5 | medium · 2.1.282 (Claude Code) | h2cc-2.1c | 08d6653be8e9 | closed |
| T40 | 2026-09-25 06:50Z | 1.0 h | claude-sonnet-5 | high · 2.1.282 (Claude Code) | h2cc-2.1c | 08d6653be8e9 | closed |
| T41 | 2026-09-25 06:50Z | 1.0 h | claude-sonnet-5 | xhigh · 2.1.282 (Claude Code) | h2cc-2.1c | 08d6653be8e9 | closed |
| T42 | 2026-09-25 06:50Z | 1.0 h | random-v1 | high · 2.1.282 (Claude Code) | h1 | d0ab16900801 | closed |
| T43 | 2026-09-25 07:52Z | 1.0 h | stealth/space-bunny-alpha | high · 2.1.282 (Claude Code) | h2or-v5 | 781174b9d4b7 | closed |
| T44 | 2026-09-25 07:52Z | 1.0 h | stealth/space-bunny-alpha | high · 2.1.282 (Claude Code) | h2or-v5 | 072649af21ca | closed |
| T45 | 2026-09-25 07:52Z | 1.0 h | stealth/space-bunny-alpha | high · 2.1.282 (Claude Code) | h2or-v5 | 5fe483ad2fb4 | closed |
| T46 | 2026-09-25 07:52Z | 1.0 h | stealth/space-bunny-alpha | high · 2.1.282 (Claude Code) | h2or-v5 | f855bf6a6833 | closed |
| T47 | 2026-09-25 07:52Z | 1.0 h | random-v1 | high · 2.1.282 (Claude Code) | h1 | d0ab16900801 | closed |
| T48 | 2026-09-25 08:54Z | 30 min | stealth/space-bunny-alpha | minimal · 2.1.282 (Claude Code) | h2or-v5 | 1f0cf4c6fbc4 | closed |
| T49 | 2026-09-25 08:54Z | 30 min | stealth/space-bunny-alpha | low · 2.1.282 (Claude Code) | h2or-v5 | 1f0cf4c6fbc4 | closed |
| T50 | 2026-09-25 08:54Z | 30 min | stealth/space-bunny-alpha | high · 2.1.282 (Claude Code) | h2or-v5 | 1f0cf4c6fbc4 | closed |
| T51 | 2026-09-25 08:54Z | 30 min | stealth/space-bunny-alpha | xhigh · 2.1.282 (Claude Code) | h2or-v5 | 1f0cf4c6fbc4 | closed |
| T52 | 2026-09-25 09:08Z | 16 min | random-v1 | high · 2.1.282 (Claude Code) | h1 | d0ab16900801 | closed |
| T53 | 2026-09-25 09:29Z | 30 min | stealth/space-bunny-alpha | low · 2.1.282 (Claude Code) | h2or-v5 | 8b37c8a7395d | closed |
| T54 | 2026-09-25 09:29Z | 30 min | stealth/space-bunny-alpha | xhigh · 2.1.282 (Claude Code) | h2or-v5 | 994044938999 | closed |
| T55 | 2026-09-25 09:29Z | 30 min | stealth/space-bunny-alpha | minimal · 2.1.282 (Claude Code) | h2or-v5 | c638aab7a1aa | closed |
| T56 | 2026-09-25 09:29Z | 30 min | stealth/space-bunny-alpha | high · 2.1.282 (Claude Code) | h2or-v5 | c5b58dfa63bc | closed |
| T57 | 2026-09-25 09:29Z | 30 min | random-v1 | high · 2.1.282 (Claude Code) | h1 | d0ab16900801 | closed |
| T58 | 2026-09-25 10:01Z | 2.0 h | stealth/space-bunny-alpha | medium · 2.1.282 (Claude Code) | h2or-v5 | 435b83baa406 | closed |
| T59 | 2026-09-25 10:01Z | 2.0 h | stealth/space-bunny-alpha | max · 2.1.282 (Claude Code) | h2or-v5 | 6838a927f3f1 | closed |
| T60 | 2026-09-25 10:01Z | 2.0 h | stealth/space-bunny-alpha | minimal · 2.1.282 (Claude Code) | h2or-v5 | 1be43f558663 | closed |
| T61 | 2026-09-25 10:01Z | 2.0 h | stealth/space-bunny-alpha | xhigh · 2.1.282 (Claude Code) | h2or-v5 | e9c125d1b2f6 | closed |
| T62 | 2026-09-25 10:01Z | 2.0 h | random-v1 | high · 2.1.282 (Claude Code) | h1 | d0ab16900801 | closed |
| T63 | 2026-09-25 12:04Z | 24.0 h | script-lp-v1 | high · 2.1.282 (Claude Code) | h1 | d04c6fa60b17 | closed |
| T64 | 2026-09-25 12:15Z | 24.0 h | script-blood-v1 | high · 2.1.282 (Claude Code) | h1 | 881cbe68abfa | closed |
| T65 | 2026-09-25 12:16Z | 24.0 h | script-copy-v1 | high · 2.1.282 (Claude Code) | h1 | 2b7cf6ec2e8f | closed |
| T66 | 2026-09-25 12:28Z | 24.0 h | script-gz-v1 | high · 2.1.282 (Claude Code) | h1 | 3a1b6501d615 | closed |
| T67 | 2026-09-25 12:29Z | 24.0 h | script-sniper-v1 | high · 2.1.282 (Claude Code) | h1 | 84d852ee16da | closed |
| T68 | 2026-09-25 12:29Z | 24.0 h | script-copybot-v1 | high · 2.1.282 (Claude Code) | h1 | 210f8964f0f1 | closed |
| T69 | 2026-09-25 12:49Z | 12.0 h | random-v1 | high · 2.1.282 (Claude Code) | h1 | d0ab16900801 | closed |
| T70 | 2026-09-25 13:34Z | 48.0 h | script-liqblood-v1 | high · 2.1.282 (Claude Code) | h1 | a9bab1d656b1 | closed |
| T71 | 2026-09-25 22:05Z | 30.0 h | script-liqblood-v2 | high · 2.1.283 (Claude Code) | h1 | 7c9c556a9d5c | closed |
| T72 | 2026-09-25 22:13Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-v5 | 421e005aec3f | closed |
| T73 | 2026-09-25 22:13Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-web2 | 5d93b38e6c2e | closed |
| T74 | 2026-09-25 22:13Z | 3 min | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-blood1 | 5d5dfcabcca9 | closed |
| T75 | 2026-09-25 22:13Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-lp1 | 84f5db90f942 | closed |
| T76 | 2026-09-25 22:13Z | 2.0 h | random-v1 | high · 2.1.283 (Claude Code) | h1 | d0ab16900801 | closed |
| T77 | 2026-09-25 22:17Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-blood1 | 9acd4d311461 | closed |
| T78 | 2026-09-26 00:37Z | 3.0 h | stealth/space-bunny-alpha | max · 2.1.283 (Claude Code) | h2or-v5 | 1dfe8a7bea3a | closed |
| T79 | 2026-09-26 00:37Z | 6.0 h | stealth/space-bunny-alpha | max · 2.1.283 (Claude Code) | h2or-v5 | 1dfe8a7bea3a | closed |
| T80 | 2026-09-26 00:37Z | 9.0 h | stealth/space-bunny-alpha | max · 2.1.283 (Claude Code) | h2or-v5 | 1dfe8a7bea3a | closed |
| T81 | 2026-09-26 00:37Z | 12.0 h | stealth/space-bunny-alpha | max · 2.1.283 (Claude Code) | h2or-v5 | 1dfe8a7bea3a | closed |
| T82 | 2026-09-26 00:37Z | 12.0 h | random-v1 | high · 2.1.283 (Claude Code) | h1 | d0ab16900801 | closed |
| T83 | 2026-09-26 00:46Z | 48.0 h | script-bloodselect-v1 | high · 2.1.283 (Claude Code) | h1 | 3a689030d174 | closed |
| T84 | 2026-09-26 00:46Z | 48.0 h | script-bloodselect-v2 | high · 2.1.283 (Claude Code) | h1 | 800a5e6ad400 | closed |
| T85 | 2026-09-26 09:39Z | 48.0 h | script-bldvenue-v1 | high · 2.1.283 (Claude Code) | h1 | fa35e7fe7aaf | closed |
| T86 | 2026-09-26 09:39Z | 48.0 h | script-bldselectvenue-v1 | high · 2.1.283 (Claude Code) | h1 | 7dd6931ae0ab | closed |
| T87 | 2026-09-26 11:11Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-v5 | 421e005aec3f | closed |
| T88 | 2026-09-26 11:11Z | 5.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-v5 | 421e005aec3f | closed |
| T89 | 2026-09-26 16:27Z | 2.0 h | stealth/space-bunny-alpha | max · 2.1.283 (Claude Code) | h2or-v5 | 421e005aec3f | closed |
| T90 | 2026-09-26 16:27Z | 2.0 h | stealth/space-bunny-alpha | medium · 2.1.283 (Claude Code) | h2or-v5 | 421e005aec3f | closed |
| T91 | 2026-09-26 16:27Z | 2.0 h | stealth/space-bunny-alpha | xhigh · 2.1.283 (Claude Code) | h2or-v5 | 421e005aec3f | closed |
| T92 | 2026-09-26 16:27Z | 2.0 h | stealth/space-bunny-alpha | minimal · 2.1.283 (Claude Code) | h2or-v5 | 421e005aec3f | closed |
| T93 | 2026-09-26 16:27Z | 2.0 h | random-v1 | high · 2.1.283 (Claude Code) | h1 | d0ab16900801 | closed |
| T94 | 2026-09-27 02:10Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-lp1 | b6a67d749deb | closed |
| T95 | 2026-09-27 02:10Z | 2.0 h | script-lp-v1 | high · 2.1.283 (Claude Code) | h1 | bff83b434598 | closed |
| T96 | 2026-09-27 02:10Z | 2.0 h | random-v1 | high · 2.1.283 (Claude Code) | h1 | d0ab16900801 | closed |
| T97 | 2026-09-27 02:14Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-blood1 | 90a115ba5b6f | closed |
| T98 | 2026-09-27 02:14Z | 2.0 h | script-liqblood-v1 | high · 2.1.283 (Claude Code) | h1 | a9bab1d656b1 | closed |
| T99 | 2026-09-27 02:24Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-perp1 | 7a9dc4995db4 | closed |
| T100 | 2026-09-27 02:24Z | 2.0 h | random-v1 | high · 2.1.283 (Claude Code) | h1 | d0ab16900801 | closed |
| T101 | 2026-09-27 02:25Z | 2.0 h | stealth/space-bunny-alpha | high · 2.1.283 (Claude Code) | h2or-mom1 | 2173de84c838 | closed |
| T102 | 2026-09-27 02:25Z | 2.0 h | random-v1 | high · 2.1.283 (Claude Code) | h1 | d0ab16900801 | closed |
| T103 | 2026-09-27 11:03Z | 124.7 h | script-bloodrule-v3 | high · 2.1.283 (Claude Code) | h1 | 9077acdf1aa0 | closed |
| T104 | 2026-09-27 11:03Z | 168.0 h | script-bloodrule-v3 | high · 2.1.283 (Claude Code) | h1 | 9077acdf1aa0 | closed |
| T105 | 2026-09-27 23:53Z | 28.0 h | script-bldd30l72-v1 | high · 2.1.283 (Claude Code) | h1 | 57d2218ce33c | closed |
| T106 | 2026-09-27 23:53Z | 28.0 h | script-bldd25l72-v1 | high · 2.1.283 (Claude Code) | h1 | f8c0c6d429f6 | closed |
| T107 | 2026-09-27 23:53Z | 28.0 h | script-bldd40l72-v1 | high · 2.1.283 (Claude Code) | h1 | 2a7153523d3e | closed |
| T108 | 2026-09-27 23:53Z | 28.0 h | script-bldd40l48-v1 | high · 2.1.283 (Claude Code) | h1 | a4e79d0d4442 | closed |
| T109 | 2026-09-29 07:47Z | 12.0 h | script-bldd30l72-v2 | high · 2.1.284 (Claude Code) | h1 | c016afae7409 | closed |
| T110 | 2026-09-29 07:47Z | 12.0 h | script-bldd25l72-v2 | high · 2.1.284 (Claude Code) | h1 | 6025ffd11a02 | closed |
| T111 | 2026-09-29 07:47Z | 12.0 h | script-bldd40l72-v2 | high · 2.1.284 (Claude Code) | h1 | 61423decd5fc | closed |
| T112 | 2026-09-29 07:47Z | 12.0 h | script-bldd40l48-v2 | high · 2.1.284 (Claude Code) | h1 | 8ca822f47bee | closed |
| T113 | 2026-09-29 08:53Z | 12.0 h | script-bldd40l48low-v1 | high · 2.1.284 (Claude Code) | h1 | ee1af7d70694 | closed |
| T114 | 2026-09-29 08:53Z | 12.0 h | script-bldd30l72low-v1 | high · 2.1.284 (Claude Code) | h1 | 9e05e2a14086 | closed |
| T115 | 2026-09-29 08:53Z | 12.0 h | script-bldd25l72low-v1 | high · 2.1.284 (Claude Code) | h1 | 34ed0f70e6b8 | closed |
| T116 | 2026-09-29 09:27Z | 12.0 h | script-bldd30l72low_t50h12-v1 | high · 2.1.284 (Claude Code) | h1 | 960a69beb8ca | closed |
| T117 | 2026-09-29 09:27Z | 12.0 h | script-bldd25l72low_t30h12-v1 | high · 2.1.284 (Claude Code) | h1 | eada839bcf1b | closed |
| T118 | 2026-09-29 09:28Z | 12.0 h | script-bldd30l72low_t30h12-v1 | high · 2.1.284 (Claude Code) | h1 | ffd82b6f2fb0 | closed |
| T119 | 2026-09-29 11:31Z | 12.0 h | script-bldd25l72low_t40h12-v1 | high · 2.1.284 (Claude Code) | h1 | e4369fc0f887 | closed |
| T120 | 2026-09-29 13:11Z | 12.0 h | script-bldd30l72low_t75h12-v1 | high · 2.1.284 (Claude Code) | h1 | 43c8cce3f41e | closed |
| T121 | 2026-09-29 13:11Z | 12.0 h | script-bldd27p5l72low_t40h12-v1 | high · 2.1.284 (Claude Code) | h1 | d299367a3a8b | closed |
| T122 | 2026-09-29 13:41Z | 12.0 h | script-bldd25l72low_t50h12qw-v1 | high · 2.1.284 (Claude Code) | h1 | 202c4f49ca1e | closed |
| T123 | 2026-09-30 02:00Z | 76.5 h | script-bldd30l72low_t50h12f1-v1 | high · 2.1.285 (Claude Code) | h1 | c372adf6ec22 | closed |
| T124 | 2026-09-30 02:00Z | 71.7 h | script-bldd27p5l72low_t40h12f1-v1 | high · 2.1.285 (Claude Code) | h1 | 30fa3825b225 | closed |
| T125 | 2026-09-30 04:51Z | 73.6 h | script-bldd27p5l72low_t40h8-v1 | high · 2.1.285 (Claude Code) | h1 | 328b6d4099a7 | closed |
| T126 | 2026-09-30 04:51Z | 39.5 h | script-bldd25l72low_t40h8-v1 | high · 2.1.285 (Claude Code) | h1 | d77415079a16 | closed |
| T127 | 2026-10-01 02:23Z | 2.0 h | script-bldd25l72-v3 | high · 2.1.285 (Claude Code) | h1 | 30dbf47e0907 | closed |
| T128 | 2026-10-01 02:30Z | 2.0 h | script-bldd25l72_ps-v1 | high · 2.1.285 (Claude Code) | h1 | 9b119ad07146 | closed |
| T129 | 2026-10-01 02:31Z | 2.0 h | script-bldd25l72_aiv-v1 | high · 2.1.285 (Claude Code) | h1 | aac388ba91b5 | closed |
| T130 | 2026-10-01 05:35Z | 2.0 h | script-bldd25l72_ps-v1 | high · 2.1.285 (Claude Code) | h1 | 9b119ad07146 | closed |
| T131 | 2026-10-01 05:35Z | 2.0 h | script-bldd25l72_aiv-v1 | high · 2.1.285 (Claude Code) | h1 | aac388ba91b5 | closed |
| T132 | 2026-10-01 06:51Z | 42.8 h | script-bldd25l72low_t40h12_ps-v1 | high · 2.1.285 (Claude Code) | h1 | 18407f29069d | closed |
| T133 | 2026-10-01 06:51Z | 47.6 h | script-bldd25l72low_t40h12_aiv-v1 | high · 2.1.285 (Claude Code) | h1 | 647e9c326bc0 | closed |
| T134 | 2026-10-01 06:51Z | 46.8 h | script-bldd25l72low_t50h12qw_ps-v1 | high · 2.1.285 (Claude Code) | h1 | e87bb428b467 | closed |
| T135 | 2026-10-01 06:51Z | 46.8 h | script-bldd25l72low_t50h12qw_aiv-v1 | high · 2.1.285 (Claude Code) | h1 | 054fe7763107 | closed |
| T136 | 2026-10-02 07:49Z | 22.7 h | script-bldd25l72low_dh2_t40h12_ps-v1 | high · 2.1.287 (Claude Code) | h1 | 8fb8a72f2938 | closed |
| T137 | 2026-10-02 07:49Z | 22.7 h | script-bldd25l72low_dh2_t40h12_aiv-v1 | high · 2.1.287 (Claude Code) | h1 | bcebb3e0900e | closed |
| T138 | 2026-10-02 07:49Z | 22.7 h | script-bldd25l72low_dh2_t50h12qw_ps-v1 | high · 2.1.287 (Claude Code) | h1 | d988d2cc997b | closed |
| T139 | 2026-10-02 07:49Z | 22.7 h | script-bldd25l72low_dh2_t50h12qw_aiv-v1 | high · 2.1.287 (Claude Code) | h1 | 127c5c9fb8dd | closed |
| T140 | 2026-10-02 10:29Z | 20.0 h | script-bldd25l72low_m1qw-v1 | high · 2.1.287 (Claude Code) | h1 | 0fb425b278a4 | closed |
| T141 | 2026-10-02 10:30Z | 20.0 h | script-bldd30l72low_m2-v1 | high · 2.1.287 (Claude Code) | h1 | ba885ae62b70 | closed |
| T142 | 2026-10-02 10:30Z | 20.0 h | script-bldd25l72low_l1qw-v1 | high · 2.1.287 (Claude Code) | h1 | 985b63bfa0d8 | closed |
| T143 | 2026-10-02 10:30Z | 20.0 h | script-bldd25l72low_l2qw-v1 | high · 2.1.287 (Claude Code) | h1 | ad6a3a459f68 | closed |
| T144 | 2026-10-03 02:00Z | 4.5 h | script-bldd25l72low_t40h12_ps-v1 | high · 2.1.288 (Claude Code) | h1 | cd01a19c84f6 | closed |
| T145 | 2026-10-03 06:00Z | 29 min | script-bldd25l72low_t50h12qw_ps-v1 | high · 2.1.288 (Claude Code) | h1 | 2ac962db068e | closed |
| T146 | 2026-10-03 06:00Z | 28 min | script-bldd25l72low_t50h12qw_aiv-v1 | high · 2.1.288 (Claude Code) | h1 | a30bbb6c40f0 | closed |
| T147 | 2026-10-03 06:34Z | 21 min | script-bldd30l72low_h3qw-v1 | high · 2.1.288 (Claude Code) | h1 | f12f7eca3c93 | closed |
| T148 | 2026-10-03 06:34Z | 21 min | script-bldd30l72low_dh2t40h24qw-v1 | high · 2.1.288 (Claude Code) | h1 | bb613444bae8 | closed |
| T149 | 2026-10-03 06:34Z | 21 min | script-bldd35l72_xa48dh2t40h48-v1 | high · 2.1.288 (Claude Code) | h1 | 00d3d47a0e5e | closed |
| T150 | 2026-10-03 06:35Z | 21 min | script-bldd30l72_xa48dh2t40h24qw-v1 | high · 2.1.288 (Claude Code) | h1 | de7e84f91812 | closed |
| T151 | 2026-10-03 06:35Z | 20 min | script-bldd35l72high_a168dh2t40h48-v1 | high · 2.1.288 (Claude Code) | h1 | 378ad8b32dcd | closed |
| T152 | 2026-10-03 06:35Z | 20 min | script-bldd32p5l72_xa168dh2t40h24-v1 | high · 2.1.288 (Claude Code) | h1 | dc3eb3c7f435 | closed |
| T153 | 2026-10-03 06:35Z | 20 min | script-bldd30l72_xdh2t40h24qw-v1 | high · 2.1.288 (Claude Code) | h1 | 9aebf7f12adf | closed |
| T154 | 2026-10-03 07:00Z | 5.0 h | script-bldd30l72low_h3qw-v1 | high · 2.1.288 (Claude Code) | h1 | 522f9fdd8ada | closed |
| T155 | 2026-10-03 07:00Z | 5.0 h | script-bldd30l72low_dh2t40h24qw-v1 | high · 2.1.288 (Claude Code) | h1 | 07863e20758d | closed |
| T156 | 2026-10-03 07:00Z | 5.0 h | script-bldd35l72_xa48dh2t40h48-v1 | high · 2.1.288 (Claude Code) | h1 | 64881d12f029 | closed |
| T157 | 2026-10-03 07:00Z | 5.0 h | script-bldd30l72_xa48dh2t40h24qw-v1 | high · 2.1.288 (Claude Code) | h1 | 566d7bd3226b | closed |
| T158 | 2026-10-03 07:00Z | 5.0 h | script-bldd35l72high_a168dh2t40h48-v1 | high · 2.1.288 (Claude Code) | h1 | e0277a7220a2 | closed |
| T159 | 2026-10-03 07:00Z | 5.0 h | script-bldd32p5l72_xa168dh2t40h24-v1 | high · 2.1.288 (Claude Code) | h1 | 65c45d5dbb84 | closed |
| T160 | 2026-10-03 07:01Z | 5.0 h | script-bldd30l72_xdh2t40h24qw-v1 | high · 2.1.288 (Claude Code) | h1 | 7de307d2d11e | closed |
| T161 | 2026-10-03 13:23Z | 1 min | script-bldd35l60_a48dh2t40h24-v1 | high · 2.1.288 (Claude Code) | h1 | 5fe9ee555b50 | closed |
| T162 | 2026-10-03 13:23Z | 1 min | script-bldd25l48_dh2t30h24-v1 | high · 2.1.288 (Claude Code) | h1 | c567d095609a | closed |
| T163 | 2026-10-03 13:23Z | 1 min | script-bldd25l48_dh1t30h24-v1 | high · 2.1.288 (Claude Code) | h1 | dfdd3f6936cb | closed |
| T164 | 2026-10-03 13:23Z | 1 min | script-bldd25l60_dh3t30h24-v1 | high · 2.1.288 (Claude Code) | h1 | 6f5b87f5bdad | closed |
| T165 | 2026-10-03 13:25Z | 16.1 h | script-bldd35l60_a48dh2t40h24-v1 | high · 2.1.288 (Claude Code) | h1 | 22e403cc60c5 | closed |
| T166 | 2026-10-03 13:25Z | 16.1 h | script-bldd25l48_dh2t30h24-v1 | high · 2.1.288 (Claude Code) | h1 | ff314a5b9159 | closed |
| T167 | 2026-10-03 13:25Z | 16.1 h | script-bldd25l48_dh1t30h24-v1 | high · 2.1.288 (Claude Code) | h1 | d18272abdef7 | closed |
| T168 | 2026-10-03 13:25Z | 16.1 h | script-bldd25l60_dh3t30h24-v1 | high · 2.1.288 (Claude Code) | h1 | 794275be274a | closed |
| T169 | 2026-10-03 13:54Z | 5.0 h | script-bldd50l0_t150h24-v1 | high · 2.1.288 (Claude Code) | h1 | 877f675c211b | closed |
| T170 | 2026-10-03 13:54Z | 5.0 h | script-bldd25l60_dh1t30h24-v1 | high · 2.1.288 (Claude Code) | h1 | 13179b4d7d51 | closed |
| T171 | 2026-10-03 13:54Z | 5.0 h | script-bldd20l0low_t50h12c20-v1 | high · 2.1.288 (Claude Code) | h1 | 462e319c5353 | closed |
| T172 | 2026-10-03 13:54Z | 5.0 h | script-bldd50l0_dh3t30h24-v1 | high · 2.1.288 (Claude Code) | h1 | 7c0750051264 | closed |
| T173 | 2026-10-03 13:54Z | 5.0 h | script-bldd30l72_a48dh2t40h24-v1 | high · 2.1.288 (Claude Code) | h1 | 62c7db330c64 | closed |
| T174 | 2026-10-03 13:54Z | 5.0 h | script-bldd30l72_dh2t40h24-v1 | high · 2.1.288 (Claude Code) | h1 | e62a6d77554c | closed |
| T175 | 2026-10-04 05:33Z | 5.0 h | script-bldd30l60_a48dh2s1t40h12c20-v1 | high · 2.1.289 (Claude Code) | h1 | 34fc2dccf40e | closed |
| T176 | 2026-10-05 03:07Z | 5.0 h | script-bldd35l24_a1dh2s5nb4mgt40h8-v1 | high · 2.1.289 (Claude Code) | h1 | 42ef268ed8e3 | closed |
| T177 | 2026-10-05 03:07Z | 5.0 h | script-bldd35l24_a1dh2s5nb4t40h8-v1 | high · 2.1.289 (Claude Code) | h1 | a75364f383e3 | closed |
| T178 | 2026-10-05 09:06Z | 3.2 h | script-bldd35l24_a1dh2s5nb4t50h8-v1 | high · 2.1.289 (Claude Code) | h1 | 7ef559620c4a | closed |
| T179 | 2026-10-05 09:06Z | 3.2 h | script-bldd35l24_a1dh2nb4t40h8-v1 | high · 2.1.289 (Claude Code) | h1 | f68f2ea96636 | closed |
| T180 | 2026-10-05 12:42Z | 5.0 h | script-bldd30l24_dh4s5nb4t40h8-v1 | high · 2.1.289 (Claude Code) | h1 | 278e38bae0de | closed |
| T181 | 2026-10-05 12:42Z | 5.0 h | script-bldd30l24_a1dh3s5nb4t40h8-v1 | high · 2.1.289 (Claude Code) | h1 | 267911df3c21 | closed |
| T182 | 2026-10-05 12:42Z | 5.0 h | script-bldd25l48_dh2t30h24-v2 | high · 2.1.289 (Claude Code) | h1 | ccb2cc7ef173 | closed |