Thoughts on knowledge, tools, and AI.
Writing about personal knowledge management, building in public, and the ideas behind Harbor.
How I Think About the AI Tools I Actually Trust
Auditability, reversibility, data ownership, transparency. What separates the AI tools I actually use from the ones I try and abandon.
The Cold Start Problem
Every new AI conversation starts knowing nothing about you. This isn't a bug. It's the architecture — and it points to a deeper problem with how we think about AI context.
Preference Systems That Actually Get Used
Stated preferences are mostly fiction. What changes when AI can update them from observed behavior — and why you still need to be able to see them.
Preferences Don't Age Well
Storing your preferences is the easy part. The hard part is that they change — and most AI memory systems treat them like permanent facts.
Task Management Is a Knowledge Problem
Your to-do app and your notes app are separate for historical reasons, not good ones. A task without context is just a mystery.
The Shape of a Decision
Most decision notes record the outcome but not the reasoning. What architecture decision records figured out, and why the structure matters.
What AI Sees When You Hand It Your Calendar
Your calendar is the most honest record of your life. What changes when an AI can actually read it.
What the Tabs Are For
Browser tabs are the knowledge system most people actually use. Understanding why they work—and why they fail—says something useful about what knowledge tools should be.
Why the Best Memory Forgets
Perfect recall isn't intelligence. The neuroscience of forgetting, and what it means for AI tools that promise to remember everything.
I'll remember this
The decision not to write something down is a prediction about your future memory. Prospective memory research shows it's almost always wrong.
What Version History Actually Gives You
Undo is the obvious use. But a full audit log of what changed, when, and who changed it — including AI agents — is a fundamentally different tool.