Pourquoi les Agents IA perdent en mémoire ? | Contexte Rot, Compression et fenêtre de contexte

Join our free community: https://www.skool.com/ai-builder-2894... (+4900 members) Do you have an AI project?: https://calendly.com/benoit-ferrere11... Our Ontology platform: https://www.growthsystemes.com/ --------------------- In this video, I explain why AI agents run out of memory and can hallucinate in long contexts. I also define the term "Context," compression systems, and potential areas of memory loss for your future AI systems. --------------------- Social Media: LinkedIn:   / benoit-ferrere1   Instagram:   / benoit.ferrere   TikTok:   / benoitferrere   00:00 Introduction 00:53 Explanation of the agent loop & memory impact 02:13 Why did Context Engineering emerge? 02:58 What goes into memory 03:52 Why does the context fill up so quickly? 04:50 The context window (128k vs 1M) 05:31 Perceived intelligence (Harness) 07:06 Memory compression 10:27 Hardware limits on memory / context window 12:12 Lost in the middle & Context Rot 14:39 Where does the agent really forget? (4 distinct zones) 17:56 Conclusion