Blog
Thoughts on AI, memory, and software development.
The RAG Pipeline Nobody Told You Was Unnecessary
The aha moment - you don’t need a separate knowledge extraction pipeline - the smartest system for extracting AI knowledge is the AI that created it.
Scan vs Seek: The Architecture Gap Nobody's Talking About
There are two ways to give an AI the context it needs. The industry picked one. I think they picked wrong.
Release 0.4.1: Shaped by Your Feedback
VSCode session management, file attachments, smarter file search, and a dozen fixes — all driven by what you told us.
You Don't Need 47 Agents. You Need One That Remembers.
The agent explosion is a confession that your AI doesn't know enough. We built the agent zoo, maintained it, and then realized the zoo was the problem.
1M Tokens: It's Not the Size, It's How You Use It
Bigger context windows solve the wrong problem. 500 tokens of the right knowledge outperforms 500K tokens of raw transcript.
Your AI Coding Tool Has Amnesia
Every session, from scratch. The most capable AI models forget everything the moment the session ends. Why this is the only problem worth solving.
It Started with a Compaction Warning
How an accidental event, a personal loss, and an obsessive question led to the world's first organizational intelligence platform.