heres-the-code-training-journal-and-model-this-is-an

nikshep (@nikshepsvn)

2025-12-30 | ❤️ 224 | 🔁 12

---

here’s the code, training journal and model

this is an experiment in cross-market state fusion: teaching an RL agent to trade polymarket by fusing binance order flow with PM microstructure

early results show increasing pnl and signs of learning

https://github.com/humanplane/cross-market-state-fusion https://x.com/nikshepsvn/status/2006020259652772269/photo/1

🔗 원본 링크

• https://github.com/humanplane/cross-market-state-fusion
• https://x.com/nikshepsvn/status/2006020259652772269/photo/1

미디어

---

🔗 Related

• 53m-total-gain-in-just-3-months-forget-gabagool22-this-is
• claude-code-워크플로우-최적화-가이드-아티클을-쓰신-분이-작년에-해커톤에서-우승한-경험
• 1x-world-model-from-video-to-action-a-new-way-robots-learn
• chain-of-view-makes-vision-language-models-move-through-a
• claude-cowork-활용-방식이야-개인마다-입맛에-맞게-anthropic에서-제안한-usecases

인용 트윗

nikshep (@nikshepsvn)

training an RL agent to trade polymarket’s 15-min crypto markets for fun

input: binance order flow + polymarket books (18 features)

output: 7 actions (hold, buy/sell × 3 sizes)

learns when to trade AND how much to bet — uses PPO on MLX, trains live, ~34 min cycles https://t.co/E0nSmXZevq

원본 트윗

🎬 영상

Tags

AI-ML LLM Crypto