Insights / Topics
Working notes from building a personal AI substrate. The durable memory, structured context, and provenance layer that turns raw model calls into a system you can actually act on. Lessons from running it daily, not from a whitepaper.
AI is becoming the substrate underneath ordinary work, not a feature bolted onto it. The throughline here is what changes when building software no longer requires being a programmer: someone who walked away from code for 25 years can ship a working tool in the time it takes to eat lunch. The shift moves the bottleneck from technical skill to knowing what is worth making, and treats agentic AI as a default capability rather than a specialist's preserve.
4 Jun 2026 · 3 min read
I run my AI advisory work on a private operating system I built myself — and actually using one of these "second brain" systems has taught me that stored volume matters far less than timing. The real step change isn't buying a generic tool; it's building a system around how you actually think and work.
15 Jan 2026 · 3 min read
I taught myself to code as a teenager, then spent 25 years away from it. This week I tried Claude Cowork — agentic AI for non-coders — and shipped a working Chrome extension over lunch.