I Thought of This When I Was Eight

Kid logic: take a computer, have it try one character of code. Run it. Fails. Try the next. Keep going, add characters, eventually something doesn’t fail — mark it, keep building. A million computers doing this in parallel.

Monkey-with-a-typewriter, but pointed at software itself.

I thought I’d invented something. I hadn’t. I was gesturing at a shape that already existed and was already hard — theory, datasets, training pipelines, an understanding of what “works” even means. None of that was in my head at eight. I just noticed that brute force was theoretically possible and felt clever about it.

While I was having that thought, Rumelhart and Hinton were publishing the backpropagation paper (1986) — the foundational piece that made neural nets actually learnable. A few years later John Koza formalized Genetic Programming (1992): computers evolving programs through selection and mutation. Real math, real results. I had a hunch. They had papers.

Thirty-plus years later, the foundations got built out properly. And now there are systems that rhyme with what I was imagining — except real, except grounded, except the result of decades of people doing the actual work.

The hard part was everything I didn’t know to think about.