Google has developed a system called WikiSkill that gives AI agents the ability to retain and apply knowledge from past tasks, according to a report by The Decoder.
What Happened
The company reports that WikiSkill functions as a persistent memory layer for autonomous agents. Rather than starting each task without context, agents equipped with WikiSkill can draw on previously learned information about both successes and failures. This wiki-style knowledge base is designed to accumulate experience over time, allowing the system to avoid repeating mistakes while building on past wins.
Why It Matters
Current AI agents often struggle with continuity across sessions—they may fail at a task, then attempt it again moments later without learning from the failure. WikiSkill addresses this by giving agents a shared repository of experiential knowledge they can reference. For developers building agentic systems, this could reduce error rates and improve efficiency in multi-step workflows.
The Bottom Line
WikiSkill represents Google's approach to solving persistent memory in autonomous AI systems, though details about implementation and broader availability remain limited.