Right! GraphRAG, VectorDBs, larger context windows, MCP servers, "tools", "skills", and now "using the cli as a user" still havent really solved some of the inherent flaws of even the latest frontier models. SLMs and fining tuning gives me hope on addressing the learning part of machine learning. At least a little more
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments