Why I care about how AI gets built
I build software for a living, and for most of my career the cost of a bad release was a broken checkout flow or an angry customer. AI changes that math. The systems shipping right now will shape how people get hired, diagnosed, taught, and policed. The choices made in the next few years will stick around long after the people who made them have moved on.
Responsible AI gets talked about as a technical problem, and a lot of it is. But most of the hard parts are human. Fairness, transparency, respecting people's ability to say no – none of these fall out of a benchmark score. You have to decide to care about them, and then keep caring when they get in the way of a deadline.
The models available today do things that were science fiction five years ago. That is exciting, and it also means we owe people some humility. If you are shipping a system like this, you should be asking who it could hurt, which biases it could quietly reinforce, and what happens when it is wrong. Then test for those things instead of assuming the answer.
As AI ends up inside infrastructure and decision-making, reliability stops being a product concern and becomes a public one. The risks range from the near and boring – privacy leaks, manipulation, confident nonsense – to longer-range questions about who holds power. No single discipline has the answer. Engineers, ethicists, policymakers, and the people on the receiving end all need a seat, and that is slower than most companies would like.
One quick note on framing – this is not a race to build the most capable thing. It is a question of what the thing is for. Moving the focus from what AI can do to what it should do sounds soft, but it changes real decisions about what gets built and what gets shipped. My read is that this is the whole game. The rest of this site is me working through it in public.
Musings
The Alignment Problem
How do you get a system to want what you meant, not what you said?
Governing AI
Who decides the rules, and what happens when the rules lag the tech.
What a Good AI Future Looks Like
The upside is real. So is the fine print.
AI in Education
A tutor for every kid sounds great. The tradeoffs are in the details.
AI in Software Engineering
AI is writing more of our code. Notes from the inside.