Bitget 黑客松用 Jev 做交易决策 Agent

Crypto T @CryptoCT01原推发布于 2026.09.21

在 X 上打开原推

🚀 Two AI agents, one mission — my Bitget AI Hackathon S2 build Competing in Bitget AI Hackathon S2 ($50,000 USDT) with two separate submissions on the Agentic Trading track — each filed under a different sub-theme: 🌉 Crossfire → Cross-Asset Execution Agent ⚡Jev Pulse → Open Theme Two distinct builds, not the same idea twice. Both paper-traded — honest logs over pretend capital. 🌉 Crossfire · Cross-Asset Execution Agent A Dual Book: crypto perpetuals and tokenized US stock contracts in one decision engine. The agent reads live Bitget marks, funding, open interest, sentiment and macro feeds, then walks mandatory rules before acting — nothing invented, unwired feeds stay marked unknown. A risk cage and equity/drawdown theatre keep the blast radius visible. Problem solved: cross-asset traders run crypto and tokenized equities in silos with separate risk. Crossfire gives them one book, one risk spine. ⚡ Jev Pulse · Open Theme Inspired by Bitget's own Jev example 👉 https://x.com/Bitget_AI/status/2101922940971516406?s=20 Jev Pulse puts TypeSafe Jev 1.13 on Agent Hub data in a ~5s loop: it scans order book, liquidations, positioning and large prints, Jev answers four typed questions (Choice / Noul / Score — no prose) in ~0.3s, and the desk simulates the trade. Every tick is logged. Problem solved: most LLM trading builds fail because the model writes essays. Jev Pulse separates judgment from execution. 🤖 AI is the decision-maker in both — it senses, it calls, it acts under a fixed risk cage. ⏰ Tokenized US stocks don't sleep, so neither does the agent watching them — built for 24/7. 🎯 Next: tighter risk layer, richer perception, toward live-capital readiness. Demo below Full walkthrough of both desks. @Bitget_AI #BitgetHackathon #AgenticUEX @bitget

这条是引用着另一条写的：x.com/bitget_ai/status/2097230641452785752