How Tickerbit works, laid open
Every holding runs weekly through the same deterministic check before the AI turns it into a view. Here is exactly what is computed, where the data comes from, and what Tickerbit deliberately does not do. None of it is hidden.
The quant layer (deterministic)
The evidence layer
On top of the quant layer come current sources: web news (Tavily), news and earnings (Finnhub) and the companies’ own filings (SEC EDGAR). If a source is missing, the pipeline degrades cleanly instead of guessing.
The AI synthesis
From numbers and evidence a language model (Gemini 2.5 Flash weekly, 2.5 Pro in the monthly deep run) forms a structured view: an action (buy, hold, trim, sell), a conviction from 1 to 5, a one-sentence reason, and, behind the scenes, the objective conditions that would change that view. Tickerbit keeps checking those conditions for you and only speaks up when one is met.
What Tickerbit deliberately does not do
- No price forecasts. Language models are unreliable at this, Tickerbit issues no target prices.
- No market timing. It does not predict when the market rises or falls.
- No trade execution. Tickerbit advises, it never trades for you (enforced by design).
- No selling or sharing of your data. Only you see your portfolio, no one else.
- No win-only highlight reel. The track record shows the misses too, not just the hits.
Tickerbit is a research and advisory tool, not personalized investment advice in the regulatory sense and not a solicitation to buy or sell. Decisions are always yours.