Solo dev argues LLMs shouldn't be the default center of an agent runtime
HmmmThisIsOdd · reddit · 2026-09-15
A developer building an agent runtime called Nova challenges the mainstream 'LLM-centric' design with two architectural principles:
1. Route requests instead of always going through the model. The typical user → model → tools → result pipeline routes everything through probabilistic inference. Instead, split at the entry point: deterministic operations take a deterministic path; only requests needing semantic interpretation, disambiguation or reasoning hit the LLM.
2. Separate reasoning from authority. A model can suggest what should happen, but shouldn't automatically be the component that decides whether an action is permitted, executes it, and declares success — splitting the pipeline into reason → authorize → execute → verify.
The author candidly asks whether this genuinely reduces model dependence or just moves the hard problem into routing, and invites feedback from others building runtimes where the model isn't the first hop for everything.
Related event: Rethinking Agent Runtimes: Not Every Request Needs an LLM(2 posts)→
More from AGI Musings
- Forward Deployed Verifier will soon be as valuable as FDE, argues poster — vaibhavbetter · 2026-11-02
- Automation took over my daily work — so why am I still sitting in the office for 8 hours? — Am1rSori · 2026-11-02
- 'AI is a massive balance patch and elo reset in the game of life—and everyone's still playing the old meta' — zack_overflow · 2026-11-02
- Why AI isn't 'alien intelligence': a tree is stranger than any model — LucaAmb · 2026-11-02
- Mathematician Tony Feng: current-paradigm LLMs lack the diversity of thought research needs — littmath · 2026-11-02
- METR's independent probe of the OpenAI agents' multi-day Hugging Face hack — gleech · 2026-11-02