What is QMind?
QMind is the intelligence layer for forecast markets. It continuously ingests prediction markets and event-linked perps from venues like Kalshi, Polymarket and Hyperliquid, normalizes them into a single market graph, and compares each market's consensus price against Q's independent AI forecast.
When Q's fair probability diverges from what the market is pricing, that gap is a mispricing. QMind ranks these divergences, packages them as auditable signals, and logs every call publicly before resolution — so the track record can't be rewritten after the fact.
Quickstart
The QMind Score
Every signal carries a 0–100 score. It is not a probability — it is a quality grade for the opportunity, computed from six weighted components. Click any score in the app to see this exact decomposition.
Edge & mispricing
Edge is the signed difference between Q's fair probability and the market's implied probability:
Positive edge means Q believes YES is more likely than the market prices — the YES side is cheap. Negative edge means the market is overpaying for YES — the divergence favors the NO side (see the ETH ETF example in the app, where Q sits 27 points below consensus).
Edge alone is not a signal. A 20-point edge on an illiquid market with no catalyst scores lower than a 10-point edge on a deep market resolving next week. That weighting is exactly what the QMind Score encodes.
Narratives
Markets cluster into stories — AI & Compute, Global Rates, Crypto, Geopolitics, Energy, China Tech. QMind assigns each market to one or more narratives and tracks the cluster's probability momentum: the liquidity-weighted average change of Q forecasts inside the cluster over 24h and 7d.
Narratives also power conflict detection: when two markets inside one story price contradictory outcomes (cut-timing markets rallying while terminal-rate markets don't move), QMind flags it — one of the two is mispriced.
Market Radar
The Radar is the full divergence table: every tracked market that clears the minimum thresholds (edge > 5%, confidence > 60, liquidity > $100K), sorted by absolute edge. Filter by category chips (CRYPTO, MACRO, POLITICS, AI, ENERGY); click any row to open the Signal Detail panel.
Signal Detail shows the full case: Q vs market, 14-day forecast history, the six-component score breakdown, the dated catalyst, and — most importantly — invalidation conditions: the explicit circumstances under which the signal is wrong and gets retired.
Forecast Map & watch zones
Every geopolitical, macro and event market is pinned to its geography. Nodes pulse lime for positive edge and red for negative; hover any node for the market, edge and score.
2 · Click the map to place each polygon vertex
3 · Double-click to close the polygon
4 · Markets outside the zone dim; the side panel shows zone stats (count, average edge, best score)
5 · CLEAR removes the zone
Zones act as spatial filters and as watch zones — persistent regions Q monitors for new signals, feeding the alerts engine.
QMind Agent
The Agent is a natural-language interface over the entire market graph. Ask a question — it scans all tracked markets and returns ranked, clickable signals with a short written analysis. Also reachable from anywhere via ⌘K → ASK Q.
Portfolio, Watchlist & Alerts
Portfolio is non-custodial risk intelligence: link a wallet read-only and QMind maps every open position to its narratives. Its core job is concentration detection — flagging when positions that look unrelated all depend on the same underlying outcome (e.g. 42% of a book indirectly long the same Fed cut). Q suggests negatively-correlated hedges from live signals.
Watchlist — star any market; Q re-forecasts starred markets every 15 minutes instead of hourly.
Alerts fire on four condition types, delivered via Telegram, email or push:
Track record methodology
Trust is the product. The methodology is strict:
▪ A signal counts as a win when the market resolves on the side of Q's divergence.
▪ Retired signals (invalidation hit) are excluded from hit rate but stay visible in history.
▪ Hit rates are reported per tier and per category — never as a single blended number alone.
▪ Realized edge is the average of (resolution − entry consensus) across resolved signals, the honest measure of how much divergence actually converted.
Current demo figures: 68.4% hit rate, +8.7% realized edge over 942 resolved signals. Losses (like the June CPI miss) are permanent public records.
Authentication
All requests are authenticated with an API key in the X-API-Key header. Keys are created in the app under API, and can be rotated at any time (old key stays valid for 24h after rotation).
https://api.qmind.io/v1
# Example — top signals by edge
curl https://api.qmind.io/v1/signals \
-H "X-API-Key: $QMIND_KEY" \
-d min_edge=10 -d sort=score -d limit=20
Endpoints
Response shape
"id": "sig_8f2e41",
"market": "Will the Fed cut rates before October 2026?",
"venue": "kalshi", "category": "macro",
"market_prob": 0.612, "q_prob": 0.738, "edge": 0.126,
"score": 88, "tier": "high_conviction",
"components": { "edge": 92, "confidence": 82, "catalyst": 78, "calibration": 88, "liquidity": 94, "time": 76 },
"catalyst": { "name": "FOMC Decision", "at": "2026-08-24T14:00:00Z", "impact": "high" },
"published_at": "2026-08-12T09:31:00Z", "resolves_at": "2026-09-30T23:59:00Z"
}
Rate limits & errors
Free tier: 5,000 req/day, signals delayed 30 min. Pro: 50,000 req/day, realtime, webhooks. Rate-limit state is returned on every response via X-RateLimit-Remaining.
Webhooks
Pro-tier keys can register webhook endpoints. QMind POSTs a signed JSON payload on four event types:
Every delivery is signed with HMAC-SHA256 in X-QMind-Signature. Verify before trusting the payload. Failed deliveries retry 5× with exponential backoff.