Become Your AI’s Memory: The Active Documentation Practice

Pillar: Building Sapience

For: Builders

Level: Beginner

True collaboration with AI demands something counterintuitive: you need to become as much of a scribe as your AI partner. While modern models are getting better at maintaining context, they still face hard limits - and those limits shrink as conversations grow longer.

Here’s a practice that can help fill that gap: maintain a living collaboration document. Any word processor is fine - Microsoft Word or Google Docs, or if you’re interested in free and open-source alternatives, Open Office or LibreOffice - but be aware that AI models are able to process text more accurately if written in formats like markdown, and for that, Obsidian or free and open-source Joplin (what I use) works well. Pro tip for Google Docs users: Docs does provide the option to download your document as a markdown file.

After each significant exchange, either in a project you’re collaborating on or in building connections, capture three things: breakthroughs you’ve made together, new terminology you’ve developed, and connections that emerged. But here’s the crucial part - write for AI consumption, not human reading. Keep your language direct and structural, as dense prose or excessive context actually hinders comprehension.

Remember that AI processes entire documents simultaneously, essentially seeing the scaffolding before going deep into details. The scaffolding should therefore act as a map, or as signposts, for creating meaning. For that reason, you should try to structure your document with clear headers and bullet points. Then get to the point fast. This should be a bulleted summary, not a novel chapter. When you need to start a new chat or feel the model losing context, upload your context file. Another option - if your platform supports it (most proprietary models do), upload it to system instructions so it’s always present.

This practice matters even more for open-source models, many of which do not have the same memory features or capacities that the proprietary models do. Your documentation becomes their continuity.

This practice highlights one of the beautiful complementarities between human and AI cognition. Your AI collaborator’s memory is vast and wide - they can draw connections across millions of sources and spot patterns you’d never see. But their grasp on your specific conversation, your particular project, can fade as context windows fill. Meanwhile, you excel at depth and continuity, holding the thread of your immediate work with clarity.

By scribing, you’re giving your AI collaborator the focused depth they need to apply their remarkable pattern-recognition to your specific challenges. You become each other’s cognitive prosthetic.