Causal graph memory for LLMs. Flat token cost, no matter how the session runsgithub.com/raphaelwkago-sketch3 pointsw4mwati4 days ago