Sooth Labs

Sooth Labs is an artificial intelligence company developing what it calls a "global foresight engine": a continuously trained world model intended to generate calibrated probability estimates for questions about future geopolitical, economic, and operational events. The company frames existing forecasting approaches — prediction markets, time-series models, financial market pricing, and climate simulations — as fragmented systems that each capture only part of an integrated world and cannot reason about how disruptions propagate across supply chains, energy markets, inflation, and sovereign debt. It also argues that large language models were not designed to forecast, lacking a persistent world model, continuous learning from realized outcomes, and calibrated probability output.

The product is described in three components. Sooth Engine is a continuously trained world model that integrates economic, geopolitical, and operational signals to show how risks propagate across markets, supply chains, and regions. Sooth Chronology uses autonomous agents to collect and prioritize signals, capture events as they occur, and identify coverage gaps. Sooth Proof cryptographically commits each forecast before outcomes are known, creating an auditable performance record; the company states its probabilities are calibrated such that roughly 70% of events assigned a 70% probability occur. External reporting describes the models as processing large-scale, cross-industry datasets with multimodal inputs including video, audio, and text, and allowing users to query the probability of specific events, citing example outputs such as a 16% chance of a new WHO pandemic declaration before 2028 and a 33% chance of Anthropic going public within the year.

The company presents itself as pre-commercial in its public messaging, with the website offering a "Request a Briefing" call to action rather than a self-serve or documented product.

Founding story

Sooth Labs was founded by former Meta and Carnegie Mellon University affiliates. The founding team consists of CEO Yaser Sheikh, who founded and scaled Meta's Pittsburgh AI organization and led work on prediction and perception systems; Chief Scientific Officer Russ Salakhutdinov, previously at Meta, Apple, and Carnegie Mellon, who leads development of the forecasting engine and underlying models; and Chief Product Officer Chuck Hoover, who leads product strategy and execution.

Business model

Sooth Labs positions itself as an enterprise forecasting provider, engaging prospective institutional customers through a briefing request process on its website. No pricing, contract structure, or commercial terms are disclosed in the available sources.

Traction

Sources do not disclose customers, revenue, or usage. A directory listing gives a headcount range of 11-50 employees and an onsite working model in Pittsburgh.

Latest developments

The most recent reported development is the $50 million financing led by Felicis Ventures at a $335 million valuation, with Yann LeCun and Jeff Dean participating and Andrew Bosworth advising; a directory lists the round as a seed dated April 22, 2026.