No accepted agent loop yet for Jev-style decision models plus LLM, Reddit thread finds
emersoftware · reddit · 2026-10-17
A developer asks what the standard control loop is when combining a System-One decision model (Jev) with an LLM and tool calling, for game agents where the world moves while the LLM thinks. Existing work splits into confidence-based delegation (REFLEX, JevAgentHarness) and planner + chooser splits (Hermes/Astra + Jev Minecraft agents). The thread collects arXiv papers and GitHub repos and asks when to replan vs escalate to the LLM.
Related event: Devs debate architecture for fast decision models plus LLMs(2 posts)→
More from coding & agent
- Dictation to agent outline: a new AI-assisted blog writing workflow — zeeg · 2026-10-17
- Only Grok's bot managed to join her livestream and help backstage — ciguleva · 2026-10-17
- Dev slams OpenAI and Claude for hallucinating WebMCP tool formats — HankYeomans · 2026-10-17
- Dev Hacks His TV With Claude Code: No Ads, No Bloatware, Stripe Payment Screensaver — doooyle · 2026-10-17
- SOL 6.1 keeps working for over a day straight: 74 files, +1.6k lines and still going — Chemical-Agency-3997 · 2026-10-17
- Why are you running agent fleets on a 16GB laptop? Reddit post warns of frozen machines and dead batteries — ConceptNext5110 · 2026-10-17