roninazure/blackglasslab
Swarm — evolutionary AI forecasting swarm that finds mispriced YES/NO odds and generates paper trades for now. Polymarket-first, Kalshi next.
What's novel
Swarm — evolutionary AI forecasting swarm that finds mispriced YES/NO odds and generates paper trades for now. Polymarket-first, Kalshi next.
Code Analysis
10 files read · 3 roundsAn LLM-powered prediction market trading engine that uses Claude to forecast outcomes on Polymarket, applies multi-factor opportunity scoring and market quality policies, runs a second-pass skeptic review, and records all forecasts (including rejected ones) in a shadow backtesting system with adapti
Strengths
The shadow forecasting and revenue POC infrastructure is genuinely well-engineered — adaptive thresholds, execution economics with fee/slippage/spread modeling, calibration bands, and proper deduplication show real financial modeling depth. The market universe policy system with tiered classification, category caps, and malformed question detection is production-grade filtering logic.
Weaknesses
The 80kb live_runner.py monolith is a significant architectural liability that buries the core pipeline logic in a single file. The 'adversarial AI committee' framing oversells what is essentially a two-call LLM pipeline (forecast + skeptic review) with a heuristic fallback, and the evolutionary agent population system in orchestrator.py operates on a separate, less-developed code path from the main live runner.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
58
Contributors
1
Files
37
Active weeks
4
Repository
Language
Python
Stars
2
Forks
0
License
—