The short version
I set the direction, AI does much of the execution, and I own the result. The facts, opinions, and decisions on this site are mine, and so is the responsibility for every word published under my name.
Where AI helps
Writing. Most of the text on this site, the Notes included, starts as my own raw thoughts, often rough and unfiltered. AI turns them into drafts and revises the wording on my instructions. I go through every sentence, sharpen it until it says exactly what I mean, and decide what gets published.
Research and fact-checking. AI helps me find and read sources faster than I could alone. Fact-checking is my job, and AI makes me more thorough at it. Before a Note goes live, AI matches every claim to the full text of its source, and I verify the result and sign off on each claim myself. Nothing goes out just because a model said so.
Code and tools. I build my own software, from small utilities to larger systems I depend on in my daily work. I specify what they should do and how they fit together. AI coding agents write most of the code, and I test and review what they produce.
Analysis. AI is fast at the heavy lifting: cleaning data, running models, building a first version of the numbers. The judgment on assumptions and plausibility stays with me. That combination, machine speed and human judgment, is where I think the real value is.
Sensitive personal data stays out of consumer AI apps.
Why I work this way
Despite all the talk of a bubble, AI is already in daily use and already creating real value. I believe the people and companies that learn to work with it early will have a lasting edge. Refusing it because we think our skills are uniquely human looks to me like a classic case of overconfidence, one of the costliest mistakes smart people make.
Humans have a long habit of assuming we sit at the center of things, and of being slow to accept it when we don’t. I’d rather not repeat that. So I let AI do what it does well, drop the skills that no longer give me leverage, and put my time into what still needs a person: asking the right questions, judging what matters, building relationships and trust with people, and taking responsibility for the outcome.
And honestly, I find all of this exciting. I’ve always been drawn to exploring, from maps to the night sky, and I sometimes feel born too late to explore the world and too early to explore the universe. AI is the frontier my generation does get to explore, and I want to be out there early.
The part I haven’t solved
Working this closely with AI has a cost, and I feel it. An AI-heavy day is still real work, just a different kind: steering, reviewing, and making decisions nonstop. By the evening I’m often more drained than after a day without it. Prompting can feel like a slot machine: one more try, one more hit of hope that the next answer is the one. It’s tempting to let the model do the thinking, and the sheer volume of output is more than anyone can properly digest.
I don’t have a lasting fix yet. Admitting the problem is the first step. When I find something that works, I’ll add it here.
Responsibility
Whether it’s academic work, a project, or a Note on this site, my rule is the same: I’m open about how AI was involved, and I stand behind every page of the result. If something on this site is wrong, that’s on me, not on a model. If you spot a mistake, let me know.
My view on all of this keeps changing, just as the technology does. That’s why this page has a date.