Google's WikiSkill gives AI agents a persistent memory of past mistakes to sharpen future performance
Smaller models just caught up to larger ones. Google's WikiSkill gives agents persistent memory — they learn from every failure, every success.
Why it matters
Agent reliability and capability just shifted. WikiSkill lets agents build and query a persistent knowledge base across runs, turning isolated trials into cumulative learning. This is infrastructure that changes how agents scale and perform in production.
The key facts
5 to knowGoogle Research releases WikiSkill framework
Agents maintain persistent wiki-like knowledge base of failures and successes
Smaller models with WikiSkill match larger models without it
Framework enables cross-run learning instead of discarding knowledge after each execution
Performance gains scale with model size but efficiency gains favor smaller models
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Google Research has introduced WikiSkill, a framework that gives AI agents a persistent knowledge base. Instead of discarding what they learned after each run, agents document both failures and successes in a wiki-like structure and use that knowledge to get better over time. Larger models benefit…