I could use a little fuel myself

Counterpoint:

Nothing weird about it in context.

Introvert by virtue of the knitting.

Separately, contra your subjective experience, my subjective experience has been that the best code reflects human thinking, because we ultimately write code for humans, not machines, as evidenced by the fact that we write code in human (albeit artificial) languages, not machine languages.

I overall agree with you, but don't know that there's much basis that humans think in "human language" or "plain language."

So...I don't really know. In my defense, I only worked in a mathematical neuroscience lab for one summer as an undergrad about 15 years ago.

Well according to the people who both created and are currently marketing this technology, it is "incredible at coding." In fact, o3 surpasses o1 by 22.8 percentage points on SWE-Bench Verified, whatever that is.

As reported by "venturebeat.com" and cited on Wikipedia.

By the point in which you know how to ask the right questions, asses the answers, and write unit tests, it seems to me that you just do know how to write code. What other piece would you not know at that point that you could have skipped learning?

I'm a software engineer, not an academic, but I'm skeptical that LLMs are on the brink of being as capable as you're implying. It still seems safest to me for people to know how to write the code in principle, but use available tools to speed up the development process.

Psh, that's nothing

I wonder if the iterative nature of the problem solving approach plays a role. The whole "start with a skateboard, make it into a bicycle, and eventually it will be a car" thing where I suppose you're supposed to assume that if you can build a skateboard, you're on your way to building cars.

Maybe discouraged, but not disabled with certainty.

@danabra.mov