A cohort is a named set of people you compare as a group. Both cohorts are drawn from the same selected sample, so a difference between them is a difference within this sample. It is not a claim about everything these people have said.
Each dot is one statement, placed at the date it was said and the position it takes on the investigation question. The bar to the right is the shift between their first and last statement here. Both the position score and the shift are curated readings for this demo, not pipeline output. Assertiveness is an experimental text heuristic over the verbatim words, not a validated measure.
A finding is a claim about the corpus. Every one opens onto the statements that support it and the statements that cut against it, because a finding with no dissent shown is a finding nobody checked.
The question, horizon, assumptions and sensitivities below were written for this demo to show the shape of the experience. The evidence attached to each assumption is real and links to source. The Minutes has not validated any prediction, and nothing here has a track record.
Does aggregate AI infrastructure capex in 2028 exceed $2 trillion?
To 31 December 2028. Twenty-eight months from the newest statement in this corpus.
An alert rule watches the corpus and tells you when something in it changes. Build one here, then read the example it would have produced.
Saved investigations keep the people, the subject and the view you were in. Reopening one restores that state.
On this address a saved investigation is stored in this browser only. It does not reach a server, it is not shared with anyone, and it will not follow you to another device or survive clearing site data. Accounts are not built yet.
What is real: every quotation, speaker, date, topic tag and timestamped source link was read from the published pages of minutesof.com, which are built from the production database. All 34 were checked back against those pages for speaker, exact wording, date and timestamp on 5 September 2026; one was found to be a truncated fragment and was replaced with the complete published quote. The assertiveness label is an experimental text heuristic over the words shown: commitment markers minus hedge markers. It is not a validated measure. What is illustrative: the position scores, the shift figures, the three findings, the whole scenario tab and every alert this page derives. They are curated for this demo. No alert has been delivered to anyone, and no prediction here has been validated or scored by the pipeline.