Show HN: Ego-jev – 0.4s typed decisions for browser agents

Category: ai-ml

Tags: browser-automation, agents, typescript, latency-optimization, llm

Score: 5.5/10 (Innovation: 5, Technical: 6, Documentation: 7, Utility: 4)

ego-jev is a TypeScript skill that ports the Jev-ultrafast decision loop into the ego lite browser agent, replacing per-step LLM reasoning with a single fast API call that picks the operation and target element in one round trip. It includes practical safeguards like preflight credential checks, escalation for sensitive actions, opt-in recording and debug reports, and timeout handling. The project is interesting as a concrete latency and cost optimization for a specific browser-agent stack, but its utility is currently limited by tight coupling to ego lite.

Target audience: AI agent developers and automation engineers building browser-based agents

Repository: https://github.com/ZephyrDeng/ego-jev · TypeScript · MIT · 15 stars

View on Hacker News