The platform built on top of this repo is launching next week.
Three ways to use it:
-
Copy ATLAS — trade alongside the autoresearch system in this repo. Live signals, auto-execution, full transparency. Up 30% since launch.
-
Build your own — describe a strategy in plain English, backtest on 18 months of real data, deploy live.
-
Marketplace — publish your agent, earn when others copy it.
SIMONS is already live on Kalshi prediction markets. 60% win rate.
As a thank you for all the feedback, ideas, and support that helped shape this project — we're giving the GitHub community 20% off. Forever, not just the first month.
→ atlasagents.co Use code GITHUB20 at checkout. First 100 only.
ATLAS is an AI-native hedge fund system with a live execution loop, scoring, and prompt evolution.
Autoresearch is the legacy prompt-improvement loop. It scores agent output, adjusts weights, and rewrites prompts when a worse version is detected.
Core files:
agents/autonomous_loop.pyagents/autoresearch_program.mdagents/scorecard.pydata/state/autoresearch_results.tsv
Darwin v3 is the standalone gene-pool and rewrite-planning layer added beside the live loop.
Entry point:
darwin_v3/runtime.py
Runs inside:
agents/execution_loop.py
Key outputs:
data/state/judge_daily.jsondata/state/janus_daily.jsondata/state/decisions_v2.jsondata/state/decisions_v3.json
Gene pool:
darwin_v3/gene_pool.db
GRAHAM is the OTC net-net screener and diligence engine. It sources SEC-filing OTC companies, calculates NCAV, applies liquidity and shell-company filters, ranks candidates trading below 0.67x NCAV, and writes weekly diligence outputs.
Entry points:
python3 graham/run.py --mode testpython3 graham/run.py --mode statuspython3 graham/run.py --mode full
Outputs:
graham/output/screener_{date}.jsongraham/output/portfolio_candidates_{date}.mdgraham/output/memos/{ticker}_{date}.md
Terminal:
/api/graham- F8 INTEL panel, GRAHAM NET-NET SCREEN section