AI bubble watch · daily check, weekdays after the US close
Loading…

Is the AI bubble popping?

A fixed framework of 33 signals, scored every trading day by an automated check that reads filings, prices, spreads and wires, and weights every source by reliability and incentive. Price breaks are noise; capex cuts, revenue misses and credit events are the confirmations.

Now

Loading the latest run…

Trend

One row per trading day. The counts are for the twenty-four stage-setting signals in pillars A to E; the amplifier is pillar F and the circularity level is pillar G, which is scored on funding events and filings rather than prices. The shape of these columns over weeks matters more than any single day.

Scoreboard

Current colors for all 33 signals. A marker shows when a signal differs from the September 15 baseline or from the previous run. Thresholds are fixed; readings refresh with each run.

Event log

Only items that moved a signal or bear on the thesis. Daily market moves live in the trend table. Hard data is a filing, price, guidance, deal or survey; opinion is a strategist, CEO or essayist; framework entries are changes to the rules themselves.

What flips the verdict

Dated events first. A move to stage 4 requires a capex cut, and the only scheduled chances for one are earnings calls.

Trackers

Four public gauges the check reads every day, kept for context. They feed signal E3, which turns yellow when two of the three composite trackers deteriorate. Values are as the sites report them; a "carried" mark means no fresh reading was found that day and the previous one was kept.

The framework: stages, thresholds, source register, changelog
The September 15 baseline: why the headlines surged, and the evidence both ways

Loading…


Data: runs.json, events.json, framework.json. Each weekday the check appends a row and commits it to this repository; the page renders from those files. Sources are tiered: only primary and wire sources can change a signal's color. Nothing here is investment advice. Repository.