System 1 agents method released for local agentic AI
Start here
Eleftherios Tzanis and Michail E. Klontzas released the code, data and results for "System 1 agents," a method for running agentic AI locally without text generation. In it, a deterministic compiler turns a tool's JSON schema and the user's request into typed multiple-choice questions, which a local decision model answers in a single forward pass; the agent then calls tools directly rather than writing out text. The authors say a CPU is sufficient. The release does not include MCP servers; the catalogues are JSON snapshots of their tool lists. Model licences vary and must be checked on each model card: OpenJev is CC-BY-NC-4.0 (non-commercial), and the Perplexity models may require accepting terms on Hugging Face.
Generated from reports · updated 22 min ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AISystem 1 Agents: a method for local, reliable and fast agentic AI
Eleftherios Tzanis and Michail E. Klontzas released the code, data and results for System 1 agents, a method in which a deterministic compiler turns a tool's JSON schema and the user's request into typed multiple-choice questions that a local decision model answers in one forward pass, so the agent calls tools without generating text.
Heat for this event
Not enough continuous observations yet to draw a trend.