The day's trade plan, before the bell.
Up to twenty setups — each with a verdict, entry, stop, targets, a 0.5–10% position size and the reasoning behind it, already worked out — plus the market-regime read that sizes them. It is what the app's Today section opens on.
300 large caps are scored by a 35-factor model across seven categories — the same model as the Stock Picker — then the pool is tilted toward the sectors the rotation model favors (×1.8) and away from the ones fading (×0.6), with no sector allowed more than 40% of it. The top seventy fan out to a validator pass — each collecting a validator verdict, a news and sentiment read, and momentum ranks measured within that day's pool — are narrowed at the downselect, and a synthesis stage publishes up to twenty setups. Only the twenty that clear the downselect get the vision model that reads the actual chart — it is the expensive step, so it never runs on all seventy. The regime call above them is rule code, not a prompt, which is why the bias label and exposure band are reproducible.
- Five slots roll from pre-market to the evening, so the briefing is never stale
- A step that fails is named rather than papered over: the reason is recorded and threaded into the synthesis prompt, which is told not to infer a read it never got
- A failed chart read is disclosed rather than hidden — a “Some analyses failed” banner shows the count on screen, so a briefing built without chart input can no longer be mistaken for a healthy one
- Equities only