Il segreto della memoria degli Agenti AI

Join my AI Academy: https://www.rizzoaiacademy.com/ Want to develop advanced AI solutions? https://inferentia.xyz IG:   / simorizzo_ai   In this video, I explain in detail how the memory of AI agents like OpenClaw, Claude Cowork, Manus AI, Genspark AI, and other advanced agent systems actually works. I start with the context problem and the context window limit, then delve into context rot, the compaction mechanism, and how agents manage memory over time to maintain consistency, performance, and operational continuity. I also discuss Agentic File Search, a new technique in which the agent directly uses the file system as external storage: an approach that could radically change the way we build AI systems and, in many cases, even replace traditional RAG. In the video, we see: why context degrades over time what context rot is how memory compaction works how an agent thinks when it needs to remember useful information why the file system can become persistent agentic memory why Agentic File Search is so important the best tools and frameworks driving this evolution Tools mentioned in the video: memU → https://github.com/NevaMind-AI/memU agentic-file-search → https://github.com/PromtEngineer/agen... Memori → https://github.com/MemoriLabs/Memori mem0 → https://github.com/mem0ai/mem0 Let me know in the comments if you think Agentic File Search can really surpass RAG or if it will remain a complementary technique in future AI systems. #openclaw #aiagent #codingagent