One number.
Four controls.
P(Bubble) is our estimated probability that the current AI investment cycle is in a bubble today. By bubble, we mean investment expectations materially exceed plausible, sustainable economic returns and depend on continued speculative capital. It is not the probability of a crash by a particular date.
Primary industry data
We begin with filings, earnings disclosures and measurable operating data: AI-attributable revenue, capital expenditure, margins, cash flow, backlog, capacity utilization, chip demand, power commitments, valuations and financing terms.
Verified information
Primary sources lead. Reliable financial reporting adds timely facts. Every observation is timestamped, source-graded, cross-checked, de-duplicated and time-decayed. Unsupported claims receive no weight.
Forecast intelligence
Relevant probabilities from liquid prediction markets are combined with track-recorded forecaster aggregates. We admit only clearly resolved, time-matched questions and adjust for liquidity, spread, long-shot bias and correlated markets.
Probability estimate
The model normalizes the evidence, applies versioned weights and produces a daily probability from 0% to 100%. The percentage is a structured researcher judgment, not a statistically calibrated frequency. Missing evidence lowers confidence; it never becomes certainty.
At launch, the process is researcher-operated under fixed rules. Automation and calibration checks will follow as evidence accumulates. Forecast markets inform the estimate; they never set it.
BACK TO TODAY'S NUMBER ↗