Sakana AI Researcher: Memory Harnesses Critical for Long-Running Agents, But Strategy Matters
AI Engineer · youtube · 2026-08-12
In a talk, Stefania Druga from Sakana AI shared experiments on memory harnesses for long-running research agents. She found that when all info fits in context, adding memory doesn't help and increases cost; but for long-horizon tasks, it's crucial. She proposed a write-manage-read control loop and compared recall policies: none, vector RAG, decisions ledger, and oracle. On 68 xbench questions, the ranked ledger performed best, even beating the oracle, because giving correct memory doesn't ensure use. Bad memory is expensive. She ran experiments on a local M3 Ultra.
More from AGI Musings
- AI Vibes Shift in 4 Weeks: Anthropic's Lead Evaporates into Fierce Competition — cwolferesearch · 2026-08-13
- Opinion: Reactions to AI Writing Stem from Differing Views on Language — toolstelegraph · 2026-08-13
- Predicting 5 Frontier AI Labs by 2026: OpenAI Leads, Chinese Open Weights Join — altryne · 2026-08-13
- DeepMind Researcher Tests Claude Code: AI Agents Are Solving Real Problems — mpshanahan · 2026-08-13
- Musk: Larry Page Calling Me a 'Speciesist' Led Me to Found OpenAI — SucceededMind · 2026-08-13
- Looking Back at Google I/O 2018: The Agentic Vision Becomes Reality — yunta_tsai · 2026-08-13