What AI Evolution is
Markets change, and a system that worked in one regime can slowly stop working in the next. AI Evolution watches for that: a deterministic drift detector notices regime change or behavior degradation, and when it fires, the AI proposes a new version of the system — v1 → v2 — using the same optimizer that improves running systems.
The rule: AI recommends, human confirms
This is a hard line on the platform. Evolution never mutates a live system on its own. It hands you a proposed v2; you compare v1 vs v2 in a backtest and decide. Nothing auto-applies, and the money-path is untouched until you approve.
When to use it
- •A system that used to perform is drifting — instead of scrapping it, evolve it deliberately.
- •The market has clearly changed regime and your design was built for the old one.
The pitfall
Rubber-stamping v2. A proposal is a hypothesis, not an upgrade — compare it honestly on the same data and reject it if it doesn’t clearly beat v1. Evolution is a decision you make, not an automatic improvement you accept. (Flag-gated; an advanced capability.)
Next
To test structural variants side by side, use Branches, A/B & Compare; prove the winner in Test Lab & Proof.