Inside the workspace — four products, one seat
Quant research, quant trading, risk and compliance, and the proof layer — one shape above the portal, no feature gating inside a tier. Eight institutional domains — markets, analytics, portfolio, trading, risk, agents, data and system.
Quant Research
The whole-market data spine — roughly 14,000 names on observed day bars — with metric and plain-English screeners, fundamentals, valuation and an auditable DCF, factor and regime analytics, and AtlasQL: plain English in, a cited, re-runnable query out. Every screen the documentation lists is in every seat.
The screen documentation →Quant Trading — paper today, by design
Specialist desks from a seventeen-desk registry argue a name in public — up to eight convene per debate, always including an adversarial red team — a ten-model signal engine measures, the ML meta-layer sizes and grades, and autopilot executes against the paper book. The house verdict tries to falsify itself before it publishes, and every sealed transcript carries a re-derivable audit seal. Real-money execution ships when the public record earns it.
Watch a debate run →Risk & Compliance — and a seat never moves real money
Exposure, drawdown, component VaR and expected shortfall, correlated stress, the mandate and limit governor, pre-trade compliance and the kill switch — computed on the paper book. Every trading surface is a paper account: there is no live brokerage connection behind a seat, so no support request, no misclick and no agent can touch real funds.
The platform catalog →The Proof Layer
Provenance on every figure — real market data where the pipes are live, and an honest label everywhere else: delayed and synthetic surfaces say so, and a figure that cannot be computed renders as unavailable, never as a stand-in. Verdicts seal to a tamper-evident chain any browser can re-derive, and the per-capability register is public.
The data-status register →