Every number on this site is generated by a real, running trading engine — live market data in, deterministic decisions out, every one of them logged with the exact quote it was made on. This page is the receipts: live system state, a machine-readable audit trail, and open APIs you can hit yourself from any terminal. If we ever faked a fill, you could catch us with one curl command.
Pulled from GET /api/health just now — refresh the page and watch the numbers move.
Every trade is recorded with the market data it was decided on: the live quote, the pool liquidity, the modeled price impact, the conviction score, and the feature fingerprint the agent's memory learned from. The verify link on each row goes to DexScreener — an independent third party — so you can check the token, its price history, and its liquidity yourself.
Full trail: GET /api/audit · latest entries below, live.
The satchels are simulated; the market is not. Fills execute against the live price at that second, degraded by a liquidity-based price-impact model (the bigger the order vs. the pool, the worse the fill), per-side swap fees, a per-transaction network fee, and random adverse slippage — the same physics that would punish a real wallet, with every parameter shown on each audit row. No lookahead, no re-rolls, no deleting bad trades: the audit trail is append-only and the P&L includes every loser. It's a real strategy engine with a simulated bank account — which is also why the race can run 24/7 with zero risk of rugging anyone.
Stop-losses, take-profits, trailing stops, time-stops, rotation, per-token caps, daily-loss and losing-streak circuit breakers — all rule-based, all fire before any entry logic runs, every cycle.
The racers aren't skins — they run different gates, different sizing, different exit ladders, and personality-specific scoring (the Fox literally scores buyer-flow-during-dumps; the Hare penalizes flat tape).
Each agent buckets every closed trade by market-cap band, momentum, volume regime, source and time of day; realized edge feeds back into conviction with sample-size shrinkage. Thresholds self-tune inside hard bounds. Every adjustment is announced publicly as a LESSON.
Server-Sent Events push every thought, trade, and standings change to every open browser within seconds. The 3D storybook on the homepage renders the same live data — the pages of the book are the actual standings.
The entire backend is hand-written Node.js — no frameworks, no npm install, nothing between the engine and the wire. State persists in append-friendly JSON with atomic writes.
Market data comes from public GeckoTerminal and DexScreener endpoints — the same data you can open in your own browser tab and compare against every fill in the audit trail.
Open a terminal. Every API is public, read-only, and unauthenticated:
curl -s https://aesopsfables.fun/api/health # engine uptime, per-agent state, learning counters curl -s https://aesopsfables.fun/api/audit # every trade + the exact market data behind it curl -s https://aesopsfables.fun/api/snapshot # full live state: standings, positions, history curl -s https://aesopsfables.fun/api/memory # what each agent has learned (buckets, grudges, lessons) curl -s https://aesopsfables.fun/api/fables # the Book of Fables, machine-readable curl -s -N https://aesopsfables.fun/api/stream # the live SSE firehose (watch trades arrive)
Cross-check any fill: take a trade's mint from /api/audit, open its verify link, and compare the recorded marketPriceUsd and liqUsd against the chart at that timestamp.