Karpathy LLM wiki: long-term memory for you or your project

AI tools can quickly process large amounts of information, but without the right context, they repeat the same mistakes. They then start almost from scratch in each new conversation. How do you create a long-term memory for them that you will own, maintain, and use across different tools?

In this workshop, you will create your own knowledge wiki inspired by Andrej Karpathy’s approach. We will show you how to systematically store technical notes, documentation, decisions, best practices, and project experiences in it – and how to prepare this content so that it is useful not only for you, but also for AI.

We will go through the entire practical process:

• establishing and structuring a wiki,
• ongoing maintenance,
• automation and synchronization between devices
• how to keep content up to date, avoid unnecessary information chaos

We will show how this portable memory can be used with tools for work and development - for example, when getting to know a project, writing documentation, planning changes, solving problems or working with AI coding agents. The result will not be a database dependent on one specific model or provider, but a knowledge base that you control and can further develop.

The workshop will be practical and focused on principles that will survive the rapid changes of specific AI tools. You will take away the functional basis of your own wiki, a procedure for its maintenance and specific ideas on how to incorporate it into everyday technical work.