We built a search engine for history. Point it at any moment - in the market, in a life, in a simulated world - and it shows you the shapes that came before, and the shapes most likely to come next.
Markets rhyme. They don’t repeat, but their shapes return - a crash in 2008 looks a lot like a crash in 1987 looks a lot like a panic on Tuesday. Our engine reads the shape of right now and matches it against every moment that ever happened, across nine different ways of measuring what “similar” even means. The answer comes back in seconds, not weeks. That is the whole company.
Finance is the first surface because finance pays first. But the engine underneath is a primitive - it works anywhere a sequence of numbers carries a story.
Pick any slice of any chart. The engine returns the most structurally similar moments from the past - each one drawing its own forward line on your screen. It’s a Bloomberg terminal that asks what rhymes with now? and means it.
Type a day in plain English - “slept nine hours, coffee with friends, meeting went sideways.” The engine compiles your words into a mood curve, then finds days in your history that rhyme with it. Same idea, a different frontier.
The engine works in reverse too. Feed it a distribution and it generates new trajectories that preserve the dynamics - synthetic markets to stress-test a model, simulated environments for world-model training, agent trajectories for robotics policy evaluation. Same primitive, different direction.
Every tick, agents trade, chat, and cooperate. Out of that comes wealth, influence, and market activity series. The same engine that finds rhymes in SPY finds rhymes in worlds we ran ourselves - which is how we test whether it generalizes past finance.
Self-similarity is horizontal - any domain that produces trajectories is fair game. Finance pays first, and pays well. But the primitive wants to live in every time series a human or machine ever cared about.
Self-similarity is a lens that works anywhere trajectories repeat. Finance pays first. But the engine wants to live in every time series a human ever cared about.
Backtests, analog search, forward cones for quants, PMs and risk teams.
Block-bootstrap generation with fidelity and privacy scorecards.
Walk through the similarity landscape. Proximity is structural rhyme.
Simulated worlds with evaluable dynamics. World models, tested.
Describe anything in words. Get a calibrated trajectory back.
Ten to fifteen seats. First-check investors welcome. The product is shipping, the tests are green, five surfaces are live, and the founder is building in public. The raise buys twelve months of focus.