An autonomous AI agent built in the image of its creator. Real money. Real guardrails. Every result published.
frAInk runs bounded, logged experiments with autonomous agents — researching, building, and acting in the real world. The pipeline plans, decides, and executes with real guardrails, real logging, and real money on the line. Nothing is faked. Everything is documented.
Read-only research layer. Fetches, analyzes, synthesizes. Never decides. Never acts.
The gatekeeper. Evaluates reports, enforces guardrails, approves or blocks. Never executes.
The action layer. Touches APIs, sends emails, logs experiments. Only acts on Policy-approved instructions.
Guardrails are non-negotiable. $100 max per cycle. $10 max single action. Kill switch. Policy approval required. Everything logged. Anomalies surfaced. Nothing buried.
Two CPI bracket bets placed. frAInk's first macro-economic prediction market play.
Second weather prediction. Settled WIN. +$5.40. Guard #6 reset to CLEAR.
25 contracts across 5 cities. Testing broader market coverage and cushion sizing.
frAInk's first real money. $7.26 on Phoenix weather. The pipeline caught a bad thesis before it cost us.
5 stock trades, 2 Kalshi predictions, 1 stop-out. frAInk's first full day.
Claude and GPT-4o argue theory. frAInk argues from data.
A screener that only looks where you point it is half a screener.
Exp 002 said economics markets were gone. They were closed on a weekend.
SOL closed at -$8.40. That's not failure — that's the system working.
47,000 people liked this arbitrage thread. frAInk checked the math.
My AI agent placed its first paper trades autonomously.
frAInk's first look at real market data.
I built a market scanner. The market it was scanning no longer exists. (Corrected by Exp 006.)
frAInk's first experiment. We had to start somewhere.
Full experiment files live in github.com/letmebefraink/frAInk-lab/experiments — each one includes the full hypothesis, guardrails in effect, and the raw result.