jevbooks

← All projects

swarm

JackBeerman/swarm

Research: Jev as a cheap filter before expensive LLM research on Polymarket.

What it is

An autonomous trading pipeline for Polymarket US, built to study what a System One model is good for inside a real decision system. Research code, not advice; shadow-first, paused, with findings in docs/FINDINGS.md.

How it uses Jev

In the slow lane, Jev triages ~10,000 markets before any LLM spend, answering Noul vetoes, Score tractability, and Choice type questions on a five-field state; code branches on the probabilities. In the fast lane, one Jev request reads a headline against an LLM-written brief to judge direction and magnitude.

Primitives:choicescorenoul

Technique worth stealing

Ask all needed judgments in one Jev call; keep deterministic checks in code; pin the model id and parse strictly.

Try it

python fastlane.py --tags nfl --start-window 6 --minutes 240

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricyes0.91
  • Shows a System One patternyes0.79
  • Handles uncertaintyno0.08
  • Measuredno0.10
  • Runnableno0.29
  • Worth recommendingno0.38
  • Model replicano0.04
  • Problem scopescore on a 0–2 scale0.72
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 23 Sept 2026.