DeepMind's Prateek Jain on MatFormer: agents that dial compute up or down by task difficulty
jainprateek_ · x · 2026-09-04
In the interview 'Routing Intelligence', Prateek Jain, Distinguished Scientist at Google DeepMind, discusses his MatFormer research: today a model spends the same compute on every token, easy or hard. MatFormer is an architecture that nests smaller transformers inside a big one, letting an agent dial compute up or down depending on task difficulty.
The interview also covers two high-level approaches to making AI agents work on hard problems for long horizons — giving them enough context to hold more information, and elastically matching model scale to difficulty at the architecture level.
More from coding & agent
- PostgreSQL 19 slips to November as LLM-assisted code review floods committers with edge cases — sanjaykalra · 2026-10-21
- LangChain adds Human-in-the-Loop beta to Managed Deep Agents — LangChain · 2026-10-21
- Voxtral 3 ships a LiveKit-based voice agent SDK streaming listen-think-speak — MistralAI · 2026-10-21
- Open-sourced pixel art normals generator ships as Rust CLI and wasm library — HugoDuprez · 2026-10-20
- Eval researcher teases guide to subjective-task evals, with a curated resource list — cwolferesearch · 2026-10-20
- Eon cuts engineer onboarding from 2 days to 2 minutes with an MCP Gateway on Bedrock AgentCore — AWS ML Blog · 2026-10-20