Notes
Shorter pieces on product judgment, methods, and concepts.
The Tragedy of the Cognitive Commons. How AI Disrupts the Regeneration of Expertise
Every firm can rationally automate junior work and plan to hire experienced people later. Collectively, that strategy consumes the shared pool of experts that AI systems still depend on for substantive validation.
The Case of the Speluncean Explorers
Lon Fuller's fictional legal case is unsettling because several coherent principles produce incompatible judgments. It offers a useful lesson for any decision where more analysis cannot remove the need to choose a value and bear its cost.
Workslop. When AI Creates More Work Instead of Better Work
AI lowers the cost of producing a memo, report, or message, but it does not lower the cost of deciding whether that output deserves attention. Cheap generation can save the sender ten minutes and create hours of review for everyone else.
Rebuilding the Apprenticeship Path in the Age of AI
Entry-level cognitive work was never just cheap labor. It was how beginners learned to notice errors, absorb standards, and face problems they would not have chosen for themselves. AI is compressing that path before a replacement exists.
Attention in the Age of AI. When Everything Is Abundant, What Remains Scarce?
AI makes code, text, and prototypes abundant. The harder work moves upstream to deciding what deserves attention, and downstream to earning other people's attention without mistaking visibility for value.
Build a Local AI Job Hunter with Markdown and an Agent
A practical, copy-pasteable system for turning each job description into a structured fit decision, a tailored resume draft, interview preparation, and an application record—without a database or an automated application bot.
The List
Frontier models are no longer released as ordinary products. Access is increasingly granted through government-reviewed lists — and when access itself creates economic advantage, the rules for handing out the keys must be transparent and open to scrutiny.
You're not using AI. You're organizing it.
The default metaphor for AI is a tool — you pick it up, use it, put it down. But that breaks when you're coordinating multiple agents across days, each with different context, and the real work is keeping state coherent between them. At some point you stop being a user and start being an organizer.
Why multiple AI roles still produce one opinion
Naming four different reviewer roles in a prompt doesn't create four independent perspectives. They paraphrase each other under different names. What actually changed the output was context isolation — and that took three versions to figure out.
What a failed RAG experiment taught me about AI memory
I built a full RAG pipeline for my personal AI system. The engineering ran perfectly. The answers got worse. The problem wasn't the retrieval quality — it was what "relevance" meant in the first place.
The better AI knows you, the harder it is to surprise you
AI personalization has a shadow: the more a model understands how you think, the more it reinforces how you already think. The thing you actually need from AI — a way out of your own frame — gets harder, not easier, as personalization improves.
Agent autonomy is not the same as good collaboration
Most AI collaboration tools optimize for making the agent more autonomous. The harder problem is designing how human and agent share state and exchange control when both sides can act continuously — and making sure the human's edits are never silently overwritten.