When I reverse-engineered ChatGPT’s memory system, I found it uses pre-computed summaries injected into every prompt. But Claude’s approach is different. Through extensive experimentation, I discovered Claude uses on-demand tools and selective retrieval, a fundamentally different architecture. But how does this actually work? And how does it compare to ChatGPT’s approach?
This is the second post in a series where I reverse-engineer the memory systems of popular AI assistants. The first post focused on ChatGPT’s memory system.
Recent Stories
Taiwan war timelines might be shorter than AI timelines
TL;DR: Most AI forecasts generally assume that if a conflict over Taiwan occurs, it will largely be about AI. I think there's a decent chance for a c…
Jan 2, 2026Groq payday is ‘the outcome you dream of’ for Kitchener tech investors
Kitchener-based Garage Capital was an early investor in the AI chip maker Groq, which was just acquired by Nvidia for a reported US$20 billion. It is the biggest deal in tech history, and the local venture capital fund stands to make millions from it.
Jan 2, 2026Current and Future Applications of Artificial Intelligence in Clinical Genetics
Artificial Intelligence (AI) is transforming medicine, and its influence on clinical genetics cannot be ignored. From variant analysis and diagnostic interpretation to predictive and prognostic ...