Path 02 · 4 min read
What a verdict means — and what it doesn't
Every covered stock carries one of four labels. The label is opinionated, but it isn't advice — it doesn't know what you own, what you paid, or what you're trying to do.
These five entries together explain what the verdict is, what supports it, and where it stops.
- 01
Verdict
The publisher's standing call on a stock — Buy, Hold, Trim, or Avoid. It's an opinion, not advice.
Every stock we cover carries a single label — Buy, Hold, Trim, or Avoid — that summarises the editorial line on the evidence as of the last refresh.
Buy reads as materially favourable, Hold as balanced, Trim as mildly unfavourable with risks rising, and Avoid as materially unfavourable. The label is the same for everyone; it does not consider what you own or what you paid.
Tap a verdict pill anywhere in the app to jump to that ticker's full profile and read the supporting bull, base, and bear cases.
Live example
Buy84%Hold62%Trim71%Avoid89%Four labels, one per company. Tap any pill in the app to open the full profile.
Right next to the label is a number — and the number doesn't mean what most people assume.
02Confidence
How sure the model is in the verdict, 0% to 100%. Higher means the evidence pointed cleanly in one direction.
Confidence is a number from 0% to 100% that the verdict model attaches to its own call. A 90% Buy means the signals overwhelmingly pointed one way; a 55% Hold means the evidence was genuinely mixed.
It is not a probability of being right. Treat it as a measure of how clean the evidence was, not how likely the stock is to go up.
Live example
92%Clean evidence, one direction68%Leans, but a real counter-case41%Genuinely mixed — call is fragileUnderneath the label, three short narratives sit side by side.
03Bull / Base / Bear
Three narratives on every verdict: the optimistic case (bull), the most likely (base), and what could go wrong (bear).
Every full verdict ships with three short narratives.
The bull case is what the optimist sees — the upside if things go right. The base case is the most likely path given current evidence. The bear case is what breaks the story — the risks that could make the verdict wrong.
Reading all three is the point. A high-conviction Buy with a thin bear case is a different signal from a high-conviction Buy with a serious one.
Live example
Bull
Hyperscaler capex re-accelerates; another guide-up.
Base
Growth holds near 30%; multiple slowly reverts.
Bear
Customer concentration cracks; export controls tighten.
Each one must be traceable. Without sources, the verdict doesn't ship.
04Citation
A traceable claim → source mapping. Every verdict must point at the specific filings, prints, or data points behind it.
Verdicts are required to cite. Each claim the model makes is tied to a source (Edgar, FRED, OpenInsider, Quiver, transcript), a one-line claim, and the value pulled.
If a verdict can't cite, the pipeline doesn't ship it. This is what separates the verdict from a generic chatbot summary.
And one more layer in front, between the AI and you.
05Compliance guard
A non-AI rules layer every verdict and every chat reply passes through before reaching you. Catches advice-speak, guarantees, and portfolio references.
The compliance guard is plain TypeScript — no model, no probabilistic call. It runs as the last step of every verdict and every chat reply, live or mock. Three bright lines: it blocks second-person trade instructions ("you should buy"), it blocks portfolio / holdings / balance references (the chat can't talk about what you own because it doesn't know), and it blocks prediction or guarantee language.
When the guard catches something, it sanitises the text in place rather than failing the response — "you should buy" becomes "the publisher leans bullish", and the reader still gets an answer. The line that the guard enforces in code is the same line that lets the product ship as a publisher rather than register as an adviser; see Publisher posture.
A high-confidence Buy with a serious bear case is a different signal than a high-confidence Buy with a thin one. Reading all three cases — and the citations — is the point.
Quick check
Did it stick?
3 questions · pass at 3/3.
1. The verdict label takes into account…
2. If the verdict model can't tie a claim to a source, the verdict…
3. The compliance guard is…