AI Layer
›Overview
Overview
The part that cannot be replicated with a static rules engine.
Seven components
| # | Component | What it does |
|---|---|---|
| 1 | Catalyst Sentinel | Reads filings and news during closed-market hours, before price reacts |
| 2 | Corporate Action Extractor | Builds a structured calendar of splits, dividends, ex-dates, mergers, delistings |
| 3 | Policy Compiler | Turns plain-English rules into deterministic on-chain policy |
| 4 | Anomaly Detection | Learns each token's normal session signature and flags deviations |
| 5 | Risk Explainer | Turns raw signals into a sentence a first-time trader understands |
| 6 | Agent Reputation | Scores autonomous agents by observed behavior |
| 7 | Threshold Tuning | Recalibrates scoring weights against realized outcomes |
Load-bearing, not decorative
Reading filings at 3 a.m., extracting corporate actions from unstructured text, and translating human intent into deterministic policy are genuinely AI-shaped problems. The enforcement layer stays deterministic and auditable — which is precisely what makes the AI claim credible rather than cosmetic.
On This Page