Whisper Radar

Opening post · Project introduction · Fictional example

Fifty voices can still be one source

October 6 issue · Updated October 5, 2026 · Taylor Burton · TSCHECH Labs

A stock story can get exciting before it gets clear. Someone mentions a possible customer. Other people repeat it. Soon the claim feels familiar, even when its supporting evidence has barely changed.

That is the question behind Whisper Radar: can growing participation across distinct communities help us anticipate wider attention to a stock within 72 hours? Alongside that experiment, I want to build a public claim ledger showing the exact assertion, the evidence supporting it, and how that assessment changes over time.

Whisper Radar is still pre-build. There is no live feed, live scan, backtest, validated probability, or established investment advantage. The model, thresholds, and weights are proposed settings. Nothing here reports a stock discovery or a tested prediction.

That would describe rising observed attention and a claim with no support found in the stated search. It would not establish that the claim is false, predict a price move, or tell us whether the stock is worth buying. Accounts are not verified people, and repeated posts are not independent confirmations. If evidence appears later, the ledger should add a dated update while preserving the earlier assessment.

The ledger and the forecast have separate jobs. A forecast concerns attention to a ticker; that attention could grow because of a different story. It cannot validate the featured claim. The planned experiment will compare a small attention model with simple post-growth and account-growth baselines on unseen observations. More complexity has to earn its place.

The immediate task is source access: establish what we can consistently and permissibly observe, how much history is available, and what we can publish. Search results can help investigate a story, but cannot establish a complete discussion baseline. If only one social source is usable, the ledger will need that explicit scope, and the broader diffusion test will wait.

Once real observations exist, the plan is to publish findings with their dates and coverage limits, then return to them after the outcome window closes. Misses, corrections, and unresolved cases belong in the record. A useful ledger may survive even if the forecasting experiment finds no advantage.

Follow along if you want to see an experiment take shape, with enough of the record visible to question its conclusions. The first thing to establish is what Whisper Radar can actually see.

AI disclosure: AI assisted with drafting from the project specification. No live source data or external stock claims were reviewed for this post.