SEER

Engine: 21/23 · 9h agoAPI →Desk →

The pricing layer for what’s actually happening.

SEER reads what’s actually happening in the world and prices markets about what’s actually happening in the world. Bloomberg for normies.

Thesis

Polymarket’s original pitch was “the world’s most accurate information source.” The UI made it a casino instead of a news product. Multi-venue fragmentation, refill flow distortion, and retail dominance broke the promise. The oracle has to live above the markets, not in them.

SEER is that oracle, packaged for three audiences. Traders who want signal. Platforms that want pricing. Normal people who want to know what’s actually happening — without being told to put a bet on it.

My news isn’t telling me what’s actually happening. SEER is the readable version.

Pricing the language of speakers used to require a quant team. Now it takes an engine.

Three legs

Now

Proprietary trading

We trade the signal ourselves. Two engines in production: mention markets (receipts below) and central-bank communication (FOMC, validated on 43 historical meetings; ECB pattern replicates). Three repeatable pattern setups with documented entries and exits.

12–18 months

Signal licensing

Market makers, prediction-market platforms, sports books, and trading desks subscribe to SEER for fair-value probabilities. Industry conversations underway with desks and platforms across the category.

Generational

Consumer truth feed

Real-time news about what’s actually happening, readable by people who are tired of spin and don’t want to be told to bet on things. The endgame is a consumer product that replaces the news.

How SEER trades

Outcome-based EV. We don’t hedge, don’t market-make, don’t scalp in-event flow. We commit to a probability at entry based on corpus + signal + market gap, then sit. Directional, conviction-sized, no dynamic rebalance.

Live engines

Engine 1 — Mention markets

Speaker-corpus probabilities for Kalshi mention markets.→

  • ResolvedDrake Iceman · Toronto
    70%42¢ YES → ✓ YES
  • ResolvedDoorDash @ Trump Vegas rally
    87%27¢ YES → ✓ YES
  • ResolvedNetflix Q4 “subscriber”
    8%NO at 22¢ → ✓ NO

4 resolved · 2 live · pre-registered before each event

Engine 2 — Central bank communication

Verbal alpha from Powell + Lagarde transcripts. Rates and equities.→

  • LiveLive call — next FOMC presser
    Hold 84.4%n=43 · argmax-acc 74.4% · Brier −0.168
  • EdgePowell → 10y treasury (V2: macro + verbal)
    Sharpe 1.32+5.4pp · CI [0.49, 2.26]
  • EdgePowell → S&P 500 (V1: verbal-only)
    +12.9ppSharpe 0.66, n=43

Walk-forward · 1 live pre-registered call (June 17) · ECB replication on /fomc

Engine 3 — Earnings calls (lean v2)

Verbal alpha on US mega-cap earnings calls. EPS + verbal + LLM outlook + sector/ticker drift.→

  • EdgeMega-10 tech direction edge (lean v2)
    +11.9pp71.1% [57.8–84.4]
  • EdgeSharpe lift (lean v2 vs EPS baseline)
    +0.34Sharpe +0.63 / CI [+0.06, +1.24] — first stat-sig
  • EdgeCross-sector failure (honest)
    −2.7ppv0 fails outside mega-tech; v2 holds on mega-cap only

Walk-forward n=45 · 25-config search · multi-testing caveat on /earnings

Engine 4 — Trump rolling feed

Live ingest of Trump statements. Grounded vs current news. Reference-adjusted market moves.→

  • LiveToday — “TOTALLY UNACCEPTABLE” on Iran response
    Brent +1.78%close-to-close · idio +0.24% vs XLE
  • EdgeIran address Apr 6 — “Iran is decapitated”
    LMT −20pp idio vs ITAceasefire path priced in 5wk
  • EdgeShadow-backtest by format
    Address 42%[CI 21, 62] · n=24 scoreable

Cron every 6h · 26 statements · 136 claims · 45 held / 60 mixed / 29 diverged / 2 null

Cross-engine — receipts & convergence

Walk-forward direction accuracy
63.3%

1,271 resolved predictions across 5 engines. Per-engine breakdown on /desk; raw at /api/receipts.json.

Convergence right now
1 topics

Where 2+ engines align on the same topic. Plus 8 actionable predmarkets picks and 4 open FOMC calls. Topic list + signal detail behind /desk.

SEER Desk — for trading desks

The operational read. Live calls, resolutions, calibration — in one screen.

Public site is the audit trail. Desk is the live working view. Auto-refreshed from each model thread’s pipeline output, dense format, no narrative wrappers.

  • · Live calls across all three engines, with current SEER vs market
  • · Recent resolutions with hit/miss tags and short attribution
  • · Calibration snapshot (Brier, log-loss, per-class) updated when pipelines re-run
  • · “Last refresh” timestamp on every visit — you see when the data ran
Access

Free during beta. Allowlist by request. Email hello@goseer.ai with your desk + asset class.

goseer.ai/desk →

Credentials required · basic auth

Beyond the engines — methodology generalizes

Iran Escalation Cycle

Trade Outcome

I built a real-time OSINT layer tracking military logistics — RAF Fairford B-2 takeoffs, A-10 buildups, refueling-aircraft turn-arounds. Markets were pricing Iran risk off headlines while the real signal moved hours to days ahead of coverage. Earlier in the conflict, USS Tripoli and other US assets were building rather than drawing down while public messaging said the war was over and oil sat near multi-month lows. Long oil. Trade worked.

Different domain, same framework — signal first, consensus second. The methodology generalizes wherever there is (a) a structured information layer the market isn’t pricing, (b) a divergence between that layer and the consensus narrative, and (c) a tradable instrument to express the gap.

Domain
OSINT / commodities
Signal
Logistics divergence
Trade
Long oil
Outcome
Worked

Coming next

Roadmap

Roadmap
Sports props
Player-prop and in-game-mention markets across NBA, NFL, NHL, MLB, UEFA, UFC, esports broadcasts. Structural-NO logic on announcer vocabulary applied to game-state.
Macro
Central bank communication (FOMC, ECB, BoJ, BoE), official statement parsing, data-release surprise pricing.
Geopolitics
OSINT-driven pricing on conflict, election, and policy-risk markets where logistics signal moves ahead of headline narrative.
Earnings breadth
Full S&P 500 transcript corpus rolling forward each quarter, structural-NO/YES pre-mapping for every reporting window.

The infrastructure is the same across all three legs. Trading proves the model. Licensing is the SaaS business. Consumer is the generational outcome.