Why Jev-class models could become a near-free judgment primitive running on-device
signulll · x · 2026-09-19
- signulll breaks down the Jev class of models: no generative decoder loop, bounded outputs, parallel inference, and extreme compression/efficiency incentives — making them ideal for on-device deployment "soon."
- On-device, judgment could be embedded into nearly every existing interaction — notifications, email, camera, keyboard, home screen, sensors, shortcuts, accessibility — at zero marginal cost and near-instant latency.
- Many problems too expensive for an LLM can be reframed as decisions: should I surface this? Is this unusual? Does the user care? Which action is appropriate? Is this interruption worth it? Did intent change? Should the UI adapt?
- Takeaway: this effectively adds a new computational programming primitive for free.
More from Infra
- Dev builds daily-driver AI coding environment on AWS Lambda MicroVMs with S3 persistence — blaizedsouza · 2026-09-19
- Running 35B–284B MoEs at 128k Context on a 12GB A2000: Measured, Plus What Didn't Matter — HomoAgens1 · 2026-09-19
- Joseph Suarez hits SOTA in 5 hours on consumer cards in a warehouse — yacineMTB · 2026-09-19
- MiMo RL livestream: infra restarts waste 33% of compute, over $400K lost — dustinvtran · 2026-09-19
- Pedro Domingos: Opposing Data Centers Means Keeping Your Country Stupid — pmddomingos · 2026-09-19
- Pedro Domingos: OpenAI and Anthropic's real moat is their massive secured compute — pmddomingos · 2026-09-19