Comment by kypro

There isn't one.

I suppose the first was just learning about NNs and realising that this was the first time I'd ever heard a way to make computers theoretically think like humans.

Deep neural networks working in the early 2010s was another because that was a unlock for a whole new paradigm of computing.

2021 I freaked out when LLMs started to become useful at writing code. Enough so that I invested a significant amount of my net worth into Google at this point assuming that coders at about a decade left.

The day ChatGPT dropped in late November I felt physically sick. I remember the day well. I was visiting family and was like a ghost. Trying to explain that something that I was worrying about for a long-time was here was near impossible (and to long extent still is).

There was a moment in early 2025 when I realised agentic coding was becoming very competent and was solving actual problems with very minimal instructions. To this point I was worried theoretically that AI could replace human coders, but had not seen examples in practise.

Mid-2025 someone non-technical said they'd been building an app, and I was sceptical it could do what they said. I was blown away by what they were able to build.

Nov-2025 a non-technical person on my team wrote a build a product from scratch to manage a range things within our business. This was non-trivial software that would have taken a team of developers a year or two to have built previously.

In early 2026 coding agents regularly wrote code better than myself, repeatedly humbling me that the decades of effort I had to put into this craft was now more or less redundant.

At some point very soon I'd expect to see very advanced software products being one-shotted and clear evidence of recursive self-improvement.

Remember in just 3 years ChatGPT has gone from being almost useless to extremely senior programers no longer writing code. Things will change in unimaginable ways from here and you should be ready for that.