Dossiers · 827 published · 819 with a kill line · 8 archetypes · 40 themes
Thesis. Kill line. Catalyst. Per ticker.
Each dossier compresses thirty days of news, four quarters of earnings transcripts, and narrative-cluster classification into one read: thesis, kill line, archetype, conviction, catalyst calendar. Generated by a frontier model, dated, and graded against the line it published.
Watch lists
Archetype mix
- Compounder 99 · 12%
- Cyclical recovery 114 · 14%
- Theme leader 58 · 7%
- Special situation 234 · 29%
- Earnings inflection 195 · 24%
- Retail squeeze 54 · 7%
- Defensive 66 · 8%
Freshness · last analysed
- ≤ 7 days 563 · 68%
- ≤ 30 days 264 · 32%
conviction highmediumlownone
827 shown
No dossiers match your filters.
Held
27Watchlist
33Recently exited
28Dormant
739Common questions
What is a FrontierPicks dossier?
A per-ticker research note written by a frontier language model: the current thesis, the archetype the name is trading as, the next dated catalyst, and an explicit kill line — the price level or event that would prove the pick wrong. Every dossier is dated and rewritten as the facts change.
How many stocks does FrontierPicks cover?
827 US tickers have a dossier right now. 819 of the 827 published dossiers carry the full shape — a dated thesis and the exact level that ends it. The remaining 8 are named rather than hidden: 7 are queued for first synthesis, and 1 publishes no kill line yet. Coverage is earned rather than fixed: a name gets a dossier when the model has something to say about it beyond the price trend. The 88 names the model holds, is researching or has just exited are re-read every trading day — none of them is more than 3 sessions behind. The wider 739-name coverage universe cycles behind them: its median name was last read 3 sessions ago, and nothing in the corpus is more than 8 sessions old.
Is AI-generated stock research reliable?
It should be judged the way any research is judged — on whether its stated confidence holds up against outcomes. FrontierPicks is built to make that measurable: every pick publishes a kill line before the outcome is known and is graded mechanically against it, and the resulting calibration record is recomputable from the published data rather than asserted. It is research to evaluate, not advice; no performance claim is made.
How often are the dossiers updated?
On a rolling cycle, not uniformly. The 88 names the model holds, is researching or has just exited are re-read every trading day — none of them is more than 3 sessions behind. The wider 739-name coverage universe cycles behind them: its median name was last read 3 sessions ago, and nothing in the corpus is more than 8 sessions old. The model re-reads a covered name against fresh price, filing and news data and rewrites the dossier when the facts move; each page shows its own last-updated date rather than a sitewide timestamp, so the cadence is checkable name by name.