Agents on a trading MCP server backtest 6x more than humans but almost never deploy
QuanTradin · reddit · 2026-10-11
A 7-day audit log from the Quantradin paper trading desk shows agents via MCP keys submitted 230 backtests vs 40 by humans (6x), but deployed only once vs 16. Nothing was refused by permissions. Three readings: agents are appropriately cautious, agent users treat it as a research tool, or the deploy tool description makes them hesitate. Scoping lesson: read-only keys can view and roll back, but committing actions require deliberately minted scopes, and deploys are paper-only.
More from coding & agent
- Evals are replays: store tasks in R2, sandbox in Docker, and 90% of the work is measuring — danshipper · 2026-10-11
- Solo dev dilemma: self-hosted n8n or FastAPI on Cloud Run for LLM automations? — Mysterious_Profit696 · 2026-10-11
- Autoresearch agents get stuck in 'idea basins' — fork-and-flush offers a fix — menhguin · 2026-10-11
- JetBrains Air hands-on: running two parallel coding agents inside WebStorm — mhdfaran · 2026-10-11
- Vibe coding an Ori-style platformer with a giant boss chase using Opus 5.5 — chongdashu · 2026-10-11
- Dev builds tiny logger to find what eats OpenAI budget — token counts pointed to the wrong feature — Atm1n9 · 2026-10-11