Write Everything Down So Your AI Does Not Have to Guess
Investor Olaf Sakkers logs meetings and introductions in a homemade database so he and his New York colleague can work apart.

TechCrunch · Olaf Sakkers is betting he can beat the emerging-manager squeeze | TechCrunchOlaf Sakkers, a venture investor behind RedBlue, says discipline is his edge.
He publishes short essays laying out what he believes about areas like robotics, and he runs the firm on a custom database he built with Claude Code, an AI tool that helps people write software.
Into that database go relationships, meeting notes, and introductions, including who he met, what was discussed, and whom he agreed to connect.
That record lets him and his one other full-time teammate, who works from New York, work asynchronously, picking up shared work at different times without a live handoff.
One reason for all that writing is his view of AI agents, AI tools that act on your behalf. They are, he says, “useless if they just have to guess,” so everything gets written down.
Questions people ask
What exactly should I write down after a meeting?
Log the person, the date, what was actually decided or promised, and any follow-up you owe. Sakkers logs relationships, meeting notes, and introductions, so a useful minimum is who, what mattered, and next step.
Do I need to build a custom database?
You can start with a notes app, spreadsheet, or shared document you will actually search. Sakkers built a custom database, but the key feature is the same: one searchable place that another person or an AI assistant can read later.
Why does writing things down help AI so much?
AI assistants work best with specific context, and they invent details when they lack it. Written notes give them names, dates, and promises to work from, which makes summaries, reminders, and drafts more accurate.



