OMEN bundles Polymarket's AI markets into two indexes. Hold Bull when you expect the acceleration to continue, Bear when it breaks – whether the music stops or Washington moves first. Then read the crash-pressure gauge to see how much stress the wider market is pricing.
Twenty live Polymarket markets, sorted into the two stories the market tells about AI. Each side is priced as its share of the pair, so BULL + BEAR always sums to 1. Open a panel to see what's inside.
Five signal families, each normalized 0–100 against calm→stress reference levels. Three are leading (prediction markets, options skew, credit – priced before the fact) and two are confirming (volatility, equity drawdown – they rise while a crash happens, not before). Only the leading side can warn; a high confirming score means the damage is already on the tape.
The index tracks a curated set of 20 Polymarket AI markets across three themes. Constituent ids, weights, and methodology are public.
See the allocations →One-time catalysts are annualized with a constant-hazard model so a December deadline and a July deadline compare fairly. Structural questions (†) are left raw.
Trace the math →The composite line is composition-neutral: a market entering or leaving the set never creates an artificial step. The trend stays whole.
Read methodology →Constituents currently trading on Polymarket across the two indexes.
Lifetime traded volume across the indexed prediction markets.
Three leading (prediction markets, options skew, credit) and two confirming (volatility, equity drawdown).
Equal-weighted gauge reading, 0–100. Today's regime: Calm.
Twenty Polymarket markets – 11 Bull, 9 Bear (3 MKT, 6 GOV) – screened for theme relevance and tradability. Closed or resolved markets drop out of the live index automatically.
Each index is 100 × Σ(wᵢ·pᵢ) / Σwᵢ over the constituents live at each timestamp, where pᵢ is the YES midpoint. Weights are equal or lifetime-volume. Deadline-adjusted basis annualizes event markets via 1−(1−p)^(365/days); structural markets (†) stay raw.
Equal-weighted mean of five families, each scored 0–100 against fixed calm→stress reference ranges. Families – not raw signals – are equally weighted, so no family dominates just by having more sub-signals.
The lead-lag study finds no statistically significant lead of the bear index (MKT sleeve) over NVDA/SOXX (right sign, p ≈ 0.14–0.19). The bubble-burst market is partly reflexive – it resolves on the very drawdowns it prices. Treat OMEN as a market-pricing monitor, not a validated forecast.
“Popping” and “bubble” are different questions. OMEN’s markets and gauges answer how much stress is priced — they cannot say the market is wrong, which is what “bubble” means. For that the monitor carries two non-sentiment anchors: audited capex vs operating cash flow for the AI-capex filers (SEC XBRL), and LEAPS-implied 1-year tail odds from the far deeper options market as a cross-check on thin prediction books. Everything else here can be collectively euphoric or collectively scared at the same time.
Bull or Bear. The full monitor runs locally – one HTML file, one Python script, no dependencies.