Inalpha 🦊 — open-source quant agents that pick the factors working now, write full strategies, and evolve them. Every order under machine approval; the LLM never on the order path.
multi-market · self-evolving · audit-grade.
I thought lookahead bias would be a data problem. Then I let an AI write factor formulas and it suggested Ref($close, -5).
That’s tomorrow’s price.
Now the DSL rejects it at parse time. “Please don’t cheat” isn’t a constraint. 🦊
akshare lost a data source and didnt tell me. empty dataframes for 3 hours before i noticed.
moved to baostock. same columns, different backend. the orchestrator didnt blink
She doesn't time entries off a fixed indicator set.
She picks the factors that are actually working right now — ranked by rolling IC.
The fortune slip is for fun. The factor selection is audited. 🦊
I let an AI write trading strategies — actual Python files — and then run them against real market data.
The first one was a clean little moving-average crossover.
The second one tried to `import os`.
Here's what stops it. 🧵