holliplex

Actually, yes! Not Mythos, but a team did use LLMs with genetic algorithms to get a best-performing warrior: https://pub.sakana.ai/drq/

Core Wars is a programming game from the 1980s that is close to my heart. So I wrote a modern parser and runtime for it, aiming for maximal compatibility with pMARS, the canonical implementation. It compiles to webassembly, and you can play with it in your browser.

Repo: https://github.com/drivasperez/dmars Interactive playground: https://dmars.drpz.xyz/playground

I made this. Getting the morph mechanic to feel "right" and not create horrible artifacting was an interesting mix of relatively principled math and a lot of hacky tuning based on rough feel.

Please, please, please stop getting Claude to write blogposts.

Yes, I thought that was probably the case, but it stuck out to me. Sorry, I didn't mean to come across as chiding you; I think the basic point about who is and isn't "allowed" to think of themselves as a computer-person is a really interesting one, and also that you can't address this topic properly without addressing gender (and race).

It's an interesting slip, given the premise of the article, that this hypothetical child is assumed to be a boy.

I think it's mostly just that we are very good at picking up on patterns, and it's extremely noticeable that half the internet has started writing in the same voice with the same tics. If Claude were quietly posting away in 2017 I don't think anyone would think twice about its output.

I would describe myself as pretty AI-positive in software engineering, and even in technical writing, but something about seeing diagrams that are clearly generated by Nano Banana Pro immediately makes me stop reading. Weird!

I understand the frustration. But if I were looking for a job, I simply would not publicly announce to the world that I am this much of a dick.

This makes sense to me -- I find that if I ask it to generate more than, say, a paragraph of code at a time[1], I'd be much better off just writing it myself.

That said, I have issues with RSI, so every saved keystroke is extremely useful to me. I find Copilot's ability to accurately predict the next, say, ten words I'm about to type very useful.

The usual insult thrown at LLMs is that they're "fancy autocomplete". I think that's actually about right, and I'm extremely sceptical about any claims made about their critical thinking skills which I think are essentially non-existent. Where I differ with most AI-sceptics is that I think fancy autocomplete is incredibly valuable.

[1]: Let's say that's about a mid-sized function definition.

Is this supposed to be so dim in light mode? It looks like it's covered in an invisible modal's backdrop.