We've been building something at Vergex. Shipping in April. It's called Harness Liquidity Agent Framework. Quick context: OpenAI recently described how they built a million-line product where no human wrote a single line of code. Three engineers, five months. The trick wasn't the model — it was the harness. Constraints, linters, structural tests, feedback loops. Mitchell Hashimoto put it well: every time an agent screws up, you engineer a fix so it never happens again. We took that idea and pointed it at quantitative trading. In our framework, every quant scenario is its own protocol — market making, stat arb, execution algos, macro relative value, prediction market arb. Each one implements the same interface, runs independently, gets its own tests and docs. The interesting part: protocols compose. StatArb → LiquidityIntel → ExecutionAlgo → HedgeEngine That's a Gold/Silver ratio trade. The signal comes from StatArb, gets filtered through liquidity scoring, executes via adaptive VWAP, and the residual delta gets hedged with COMEX futures. Four separate protocols, wired together as a DAG. You can do the same thing across asset classes. An election probability shift on Polymarket triggers a macro relative value calculation on USD/MXN, which feeds into a cross-asset stat arb signal, gets tail-hedged, and routed through execution algos. Five protocols, one composition graph. The one thing agents can't touch: risk rules. Position limits, drawdown thresholds, circuit breakers, compliance — all hardcoded. Only humans change those, through an explicit approval process. When protocols compose, risk constraints get stricter, never looser. We're building this for US equities, FX, metals, indices, and prediction markets. Full write-up coming with the April release.(Tinkle 🔶🦞🧑‍🍳⚡)