Memory, Context, and Knowledge Deck
Distinguish context window, saved memory, retrieval, and model weights in everyday LLM use.
Context window vs. saved memory The model can only use saved memory if the product brings it into the current generation path. Objection: "But I told the model this last week." Someone expects a past chat to control today's answer. Your line Maybe, but let's check whether that fact is in this chat context, saved memory, or a retrievable system of record. Past conversation alone is not always durable knowledge. Treating chat history like a governed database. It redirects from expectation to access mechanism. What are model weights? Learned parameters created during training or fine-tuning that shape how the model…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in