Evolution Lab · live & public

An AI is rewriting real trading bots. You're watching it happen.

Every 3 hours, Grok proposes one mutation to a live paper-trading bot, we backtest it head-to-head against the original, and auto-promote only genuine winners. Win or lose, every attempt is public — with the AI's reasoning. Paper money. No fabricated numbers.

Runs every 3h · — runs total · last run — · —
▶ Live cycleWatch one evolution attempt, start to finish
1The bot Grok was handed
N
—
—
Parent P&L
—
Trades
—
Model
—
Backtest
retained history
Grok receives a concise fact sheet — current config, recent backtest stats, and the price history it'll be re-tested on. It must reply with exactly one config change and the reasoning.
2Grok thinks & mutates
Grok grok-4.3
↳ Proposed mutation · 1 change
3Head-to-head backtest
Original Mutated
——
—
The bar is high on purpose. A candidate is promoted to a new live version only if it's profitable on its own, has positive expected value, and beats its parent over ≥15 trades. Most don't. That's the selection pressure working.
Mutations proposed
—
by Grok, all-time
Genuine improvements
—
beat parent on EV
Open candidates
—
tracking forward
Promoted to live
—
cleared every bar
Strategies touched
—
in the gene pool
AI spend
—
real API cost
Winners are rare by design — we show the hunt, not just the trophies. The reasoning and the near-misses are the point.

The EvoTree

every mutation attempt, branching from the original — green ring = improved, red = no improvement, purple pulse = live
Lineage
Inspect node
Click any node in the tree to see Grok's exact mutation, the head-to-head result, and the reasoning behind it.

Every attempt, with the reasoning

newest first · the "why" is the magic

Closest calls

the near-misses — mutations that moved the needle but didn't clear every bar

Do this yourself — the playbook

the whole loop is repeatable; here's exactly how

The self-improvement loop, step by step

No magic — a fact sheet, an AI, a backtester, and a promotion bar. You can run this on your own bots.

1

Pick a bot & pull its stats

Take a live paper strategy and gather its real backtest numbers — P&L, win rate, trade count, current config.

2

Hand the AI a fact sheet

Send Grok a concise summary and ask for one config mutation plus the reasoning. Constrain it to diff-shaped output.

3

Apply exactly one change

Parse the proposed diff (e.g. yes_ask_min: 0.85 → 0.87) and build a candidate config. One variable at a time keeps cause & effect clean.

4

Backtest head-to-head

Run candidate vs. parent on the same retained price history. Identical data, identical fills — only the rule changed.

5

Apply a strict promotion bar

Promote only if the candidate is profitable, has positive EV, and beats the parent over ≥15 trades. Otherwise: log it, learn, move on.

6

Track winners forward, publish all

Promote winners to a new live version and watch them forward. Show every attempt — win or lose — with the reasoning. That's the transparency.

$1
Coming soon · the "you try it" payoff

Run this exact engine on your bot.

You just watched our system improve its own bots. Soon you'll point it at yours: for $1, Grok studies your strategy, proposes a mutation, and backtests it head-to-head — same loop, same transparency. The reasoning is yours to keep.

When the $1 tool ships — and when a bot graduates. No spam.