Know what’s coming. Know what to do.

Knowable Labs builds end-to-end forecasting infrastructure. We produce best-in-class forecasts on news, politics, and finance, then execute them as positions on prediction and capital markets, with hard risk guardrails and a full audit trail.

See the track record

Track record

Every number here is checkable. The trading curve below is live and real money, not a backtest. And the benchmark score comes from ForecastBench, which grades forecasters on contamination-free questions that resolve after model training cutoffs, so nothing can be memorized.

Live trading on Kalshi

Since inception —

Last 60 days —

Trades —

Settled —

— —

Return since inception Trade placed

Fully autonomous: the system picks the markets, sizes the positions, and executes. Every trade is logged.

ForecastBench standings

On ForecastBench’s contamination-free questions, Knowable Labs holds the top market score on the board.

Forecaster	Market score higher is better
Knowable Labs	77.1
Superforecaster median human gold standard	76.3
Gemini (Google DeepMind)	76.0
Grok (xAI)	72.9

Standings as of June 29, 2026. Verify on the public leaderboard: we run as “Anonymous 8” (tick “show anonymous teams”).

The system

One stack, two layers. A forecasting engine built to be the best in the world, and an execution engine disciplined enough to be trusted with capital. Each runs on its own, and they are built to run together.

Forecasting infrastructure

Calibrated probabilities on news, politics, and finance, produced at scale. The pipeline ingests questions, gathers evidence, scores them, and retrains continuously as events resolve, so it never goes stale. It runs at a small fraction of frontier-model cost, so it covers thousands of questions at once instead of a curated handful.

Question intake: any resolvable question, from a market feed or your own list.
Evidence gathering: live news and data pulled per question, cited in every forecast.
Calibration: probabilities scored against resolved outcomes and corrected, continuously.
Coverage: thousands of open questions tracked at once.

Execution infrastructure

Turns forecasts into positions on prediction and capital markets. Built for scrutiny, not just returns:

Risk guardrails: position and exposure limits enforced in code, not in policy documents.
Explainable audit: every trade traces back to the forecast and evidence that produced it.
Full logging: every decision recorded end to end, replayable after the fact.
Continuous training: realized outcomes feed straight back into the forecaster.

Founders

The founders spent decades shipping ML and GenAI at Google. Now they're building the forecasting and trading system they always wanted.

Yew Jin Lim

Co-founder · PhD in Machine Learning

Ex-Director of Engineering on Google Search, with 18 years building planet-scale ML and GenAI systems. PhD from the National University of Singapore.

LinkedIn →

Gautham Srinivas

Co-founder · Product leader in ML quality

PM leader on Google Search AI. Before Google, an early engineer at Flipkart and one of India's top 10 competitive programmers through the IOI training camp.