Foresight as Infrastructure

The world is being asked to make more, faster, higher-stakes decisions under greater uncertainty. The tools most leaders use to navigate that uncertainty — gut, spreadsheets, the occasional consultant deck — were built for a slower world.

To keep up, foresight must become a piece of infrastructure, not a soft skill. The teams that have it will have a compound advantage. The teams that don't will be flying blind into a faster-moving fog.

As AI accelerates the pace of execution exponentially, the need for better decision-making has only risen. Velocity without direction is just motion. Businesses live and die on the quality of their decisions.

Sooth (opens in new tab) enters the market with its own bet on the future: that the next great use for AI won’t be just summarizing more documents or writing more code. It’ll be giving businesses the tools they need to see into the future further and more clearly.

A world model for foresight

Sooth is a world model for long-horizon, cross-domain forecasting that can deliver highly specific, auditably accurate predictions.

How that works in practice: You give Sooth a time-bounded, verifiable proposition — Will U.S. CPI exceed 4% year-over-year at any point in 2026? — and it returns a calibrated probability along with a structured, causal timeline of the events driving that estimate. Then you can do the thing that no chatbot and no prediction market lets you do: change an assumption in the causal chain — assume Iran signs a ceasefire — and watch the downstream probabilities shift in real time. Scenario planning becomes interactive, auditable, and quantitative. And all of the assumptions are both editable and auditable.

Importantly, the model that Sooth runs on to deliver this intelligence isn’t an LLM. LLM’s are structurally unfit for this kind of high-precision forecasting, because they optimize for next-token prediction rather than proper probabilistic scoring rules that take into account the external world state. Instead, Sooth runs on a time series model, a fundamental step forward in the underlying model technology. This is the next leg of the race to AGI.

The idea of “risk” is predicated on the unknowability of the future. Remove that unknowability, and the entire calculus of risk changes. It’s this urgency that’s guiding Sooth’s initial market focus on financial services, defense, and national resilience: they’re the areas where the cost of being wrong is highest. But long term, we believe the opportunity for Sooth is even bigger. Every business is a risk business, and every business needs better decision-making tools.

A team to build the future they can see

For an idea this ambitious, the team can't just be very good. It has to be exceptional. Sooth's is.

Sooth was founded in January 2026 by Yaser Sheikh (opens in new tab) and Ruslan Salakhutdinov (opens in new tab). Yaser was most recently a VP at Meta and is on the Robotics faculty at Carnegie Mellon. Ruslan was VP of GenAI Research at Meta, is a CS professor at CMU, and did his PhD under Geoffrey Hinton. His research has been cited more than 200,000 times. Together, they bring exactly the combination this problem requires: deep AI research credibility (opens in new tab), the systems experience to ship infrastructure at scale, and a clear, opinionated point of view on why existing approaches fall short.

They’re a highly credentialed, visionary duo who’ve also already proven they can deliver. They showed up at our first meeting with a prototype that came close to matching the accuracy of prediction markets. As soon as we met the founders, we knew there was something special here. We committed on the spot to leading their Seed Round.

We're thrilled to be partnering with Yaser, Ruslan, and the entire Sooth team as they build the foresight layer for the AI era.