again, the AI labs are obsessed with building reasoning-native language models when they need to be building *memory-native* language models – this is possible (the techniques exist)
– no one has done it yet (no popular LLM has a built in memory module)
– door = wide open

→ View original post on X — @jxmnop,