Routing Intelligence: Prateek Jain on Long-Horizon Agents
AnneliesGamble · x · 2026-08-26
An interview with Google DeepMind Distinguished Scientist Prateek Jain on building AI agents for long-horizon tasks.
Key points:
- Two approaches: Expanding context window to hold more of the problem "in-mind", or building scaffolding (retrieval, memory, planning, tools, tests, sub-agents).
- Core challenge: Finding the right balance between context length and scaffolding complexity for efficiency and elasticity.
- Prateek co-leads Gemini's long-term innovation area, focusing on efficiency and ambitious research.
Related event: DeepMind Scientist on Long-Horizon Agents: Context vs Scaffolding(2 posts)→
More from coding & agent
- Ditched Spotify's $20/mo for $200/mo of Claude to build his own music app — djcows · 2026-10-15
- Skeleton-first prompting: Anthropic engineers show how to generate rigged Blender monsters in minutes — GCWebDesigner · 2026-10-15
- Python 3.15 ships: lazy imports, frozendict, and a faster experimental JIT — _jaydeepkarale · 2026-10-15
- "Agentic coding 10x's developers" — the code, the bugs, and the people pretending to read the diff — blaizedsouza · 2026-10-15
- OWASP publishes Non-Human Identities Top 10 for securing machine and AI agent identities — gleech · 2026-10-15
- Hot Take: Agentic Automation of Mundane IT Is a Gold Mine — and Extremely Hard — Paimaamu · 2026-10-15