How it's
built
What the board measures, where the numbers come from, and — more usefully — what they are not.
The question — one thing only: do the different factors on a ticker point the same way right now? A single bullish signal means little; six of six lining up means something. The board is built to make that agreement, or the lack of it, visible at a glance.
The engine — every read calls Grok with its x_search and web_search tools, so the model searches X posts and the open web at the moment you ask rather than recalling what it saw in training. X matters here: sentiment moves there first, and no other model reads it natively. A read takes 30–90 seconds because the search runs many rounds.
The six factors — social sentiment on X, news and catalysts, analyst and institutional flow, price momentum and technicals, fundamentals and earnings, and risk. Each is scored −100 to +100 and must cite dated evidence. Risk is scored inversely: elevated risk pushes it negative. The output is forced through a strict JSON schema, so a factor can never quietly go missing — if the model found nothing, the row says so and scores near zero.
Alignment vs. confidence — two different things, and both matter. Alignment is how much the six agree with each other. Confidence is how good the underlying evidence was. A high score with low confidence means the model found little and is guessing; high alignment with low confidence is the trap this board exists to expose.
Reuse — a read is expensive, so one is kept and reused for two hours. Searching a ticker you already looked at in that window returns the saved read instantly and calls nothing; the header says Reused and how old it is. Your history lives in this browser only. The Worker also holds each read in Cloudflare's edge cache for the same two hours, so a colleague on another machine gets it free too. Re-read ignores both and pays for a fresh one.
What it is not — there is no price feed, no backtest, no position sizing and no memory between reads. Scores are not comparable across days, because each read searches fresh sources. Treat it as a fast structured briefing on what the internet is saying, and nothing more.
For prices, filings and ratings, always go to the primary source.