Models Don't Have Agency. Systems Do. — a structural take on where agentic behavior really lives
sethjuarez · x · 2026-09-03
In an 18-minute essay, Seth Juarez argues that language models only emit tokens and never gain agency: everything "agentic" comes from a runtime that shapes those tokens into executable intent. Drawing on the philosophy of action, he defines an agent as a being whose actions arise from intention — a test the model itself fails. The key gap: ask a model to book a flight and it produces a friendly sentence a human can act on, but a program can't reliably extract destination, date and time from it. The fix isn't making the model smarter but constraining the shape of what it may say, moving from word to structure to function so the runtime can act. This structured handoff, not the model, is where agency lives.
More from coding & agent
- MongoDB Atlas now powers the virtual file system behind LangChain's Deep Agents — BraceSproul · 2026-09-04
- E2B sandbox runs RL rollouts up to 3x faster, cutting idle GPU time and training cost — badphilosopher · 2026-09-04
- Claude Code adds /limit-reset, letting users manually reset the 5-hour usage cap once per week — dotey · 2026-09-04
- LM Studio Launches Bionic, a Fully Local Agent for Work and Code — mattturck · 2026-09-04
- Perplexity launches fully local Portable Computer on Linux RTX GPUs and DGX Spark — AravSrinivas · 2026-09-04
- SpecterOps Open-Sources 79 Skills and 22 Agents for AI-Assisted Infosec Tradecraft — cyb3rops · 2026-09-04