CompoundIQ is an in-progress, paper-first trading team. Versioned hypotheses become testable strategies; only approved, enabled, unexpired signals can reach the simulator; every fill returns as feedback.
An illustration of the policy gate above — flip a condition, send a signal, and watch where it stops. No live orders, no market data; the real loop runs on paper by design.
Flip a gate, then send. The gate consumer checks all three conditions at execution time.
Illustrative only — labels describe engineering behavior, not investment performance.
Research and execution never share working state. Each role has a narrow mandate, communicates through versioned contracts, and can be stopped without confusing the rest of the team.
The research service owns hypotheses, canonical data snapshots, backtests, diagnostics, stress tests, and promotion decisions. It can propose a strategy or signal, but it cannot place an order.
A signal becomes actionable only when its strategy is approved, the deployment is enabled for the correct venue and environment, and the signal has not expired. The consumer checks all three at execution time.
The crypto service refreshes market data, consumes only gated signals, and records simulated fills with fees and slippage. It is paper-only by construction while the operating evidence is still being assembled.
Every simulated fill becomes structured feedback. Research can compare expectation with execution, identify divergence, and refine or reject the next hypothesis without sharing mutable working state with execution.
CompoundIQ separates the ability to form an opinion from the authority to act on it. Every boundary is visible, testable, and deliberately harder to cross.
Autonomy is proven in simulation first. Live execution remains disabled while research, controls, and operations mature.
Research has no broker credentials, and the browser never receives them. Execution capability stays inside the venue service.
Shared schemas and mirrored Python and TypeScript contracts land before a service relies on them, keeping each specialist independently replaceable.
The private dashboard requires an explicit confirmation before enabling or disabling a deployment and explains the consequence of the action.
These labels describe engineering progress, not investment performance.
Canonical snapshots, hypothesis tracking, backtests, market-structure diagnostics, stress evidence, and promotion gates.
Hourly ingest, research, and paper execution completed end-to-end without a manual trigger.
Strategies, signals, portfolio, and deployment controls are visible behind authenticated access.
A second venue specialist is part of the architecture but has not entered implementation yet.
No live orders. The system remains deliberately paper-only while evidence and controls compound.
Bring the decision you don't dare fully automate. We'll design the gate around it, together.
Governance works in trading. See it hold a real-estate-media delivery line together—from order email to the final gate.