Roadmap

Assimilative causal inference — filter, smoother, entropy metrics, causal influence ranges — implemented natively in Rust and run per-feature across a multi-hundred-symbol equity universe. Regime detection, transfer entropy, and the statistical validation layer run in the same core.

Strategies are generated, backtested walk-forward, placebo-tested, and evolved entirely in simulation. Failed ideas are archived with their post-mortems. Nothing leaves the loop until it survives it.

Surviving strategies run against live market data with paper execution — since March. Every order passes a nine-check risk gate; every decision lands in an append-only audit ledger. This is the proving ground for controlled execution.

Define a claim and what would prove it wrong. The system helps you sharpen a hunch into a falsifiable thesis, with the criteria fixed before any test runs.

Data and narratives confront the thesis from both sides. Every item carries its source and trust tier. When the two disagree, the divergence is surfaced as a finding of its own.

Supported, challenged, falsified, or insufficient evidence — recomputed as new information arrives. A verdict always shows its confidence and what would change it. A thesis from March that stops holding in June tells you so.

An append-only record of every thesis you have run and how it resolved: a track record of your own thinking, built up claim by claim.

Teams of agents — researcher, analyst, critic, risk assessor — each working a thesis end to end. A thousand simulations in parallel, over different theses or different readings of the same one.

Strategies compete under continuous selection. Survivors evolve and recombine; failures are archived with post-mortems. Team histories persist, so you learn which lines of reasoning have earned trust over time.

Theses constrain and inherit from each other — a macro thesis about rates shaping an equity thesis about homebuilders. Teams research the intersections. A portfolio becomes a structured argument you can inspect at any level.

A strategy that has survived simulation and held up on paper earns a small real allocation — behind position limits, loss limits, and a kill switch that is never optional.

Allocation grows only with demonstrated performance, and shrinks the moment live behavior drifts from what was validated. The same discipline we apply to ideas, applied to money.

Exposure, drawdown budgets, and correlation across strategies visible and controllable in one place — because scaling capital without scaling risk visibility is how systems blow up.