about

I build AI systems that learn from their own interactions. Not chatbots, not wrappers around existing models. The kind of systems most people say are still years away.

This blog exists because most of what I read about AI is wrong in ways that matter. Not wrong about the facts, usually. Wrong about the mechanisms. People describe what AI products do without understanding how they work, and that gap produces bad analysis.

So I write about how things actually work. When a paper makes a claim, I read the architecture. When a product says "self-learning," I ask what the learning mechanism is. When the industry says something is impossible, I check whether anyone has tried the obvious alternative.

I use AI as a writing collaborator. The ideas, analysis, and positions are mine. The AI helps me draft and refine the prose. Every article goes through my review, and I rewrite anything that doesn't sound like me or hold up under scrutiny. I'm disclosing this because a blog about AI honesty should be honest about its own process.

No newsletter, no tracking, no cookies. Just articles and an RSS feed.