alefnula

I think this frames the tools too narrowly.

If Vite, Bun and uv were just "make builds faster" projects, then maybe the returns are diminishing. But the acquisitions by Cloudflare, Anthropic and OpenAI suggest this layer is becoming more strategic, not less.

These tools sit in the software supply chain: dependency resolution, project structure, tests, builds, runtimes, deployment paths and increasingly AI-agent execution loops. They define the default path for building software, and they are where AI-generated code gets tested against real dependencies, builds, tests and deployment constraints.

So I don’t think they’ve achieved all they meaningfully can. The value is shifting from raw build speed to control over the workflow layer where software is assembled.

I haven’t used Claude Code much, but I found Codex extremely frustrating. It doesn’t pay attention to anything in AGENTS.md, it’s completely incapable of removing code and is frustratingly defensive.

If you use it, the codebase constantly grows. Even when you explicitly instruct it to remove something, you always end up with more lines of code in the project than before the instruction. Also (I used it for Python and TypeScript) the code was littered with getattr(...), .get(...), isinstance(...), and TypeScript equivalents (typeof, ...). Even though I religiously type‑annotate everything.

We live in a society that defines itself not in terms of what it is but in terms of what it's not. We are not decent and kind human beings; we are not racists or not homophobes. We can't just not care; we have to be atheists. Even with bad stuff: We are not victim blamers and choice deniers; we are anti-abortionists.

And the products marketed to us are just a reflection of what we are. Gluten-free bread. Sugar-free drinks. Non-hallucinogenic LSD. No THC weed. Low-fat milk. Etc.

Everything is becoming defined in terms of what it's not.

Can you share the link to the study you're referring to? I will soon be choosing a school for my kid, and I'm very curious to see if there are any actual differences in outcomes. All the studies I found say there are. (For example https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5670361/)

I'm surprised that no one is commenting on macros. For me, that is the most exciting feature! It'll make writing code so much more enjoyable and avoid all of the code generation happening at the moment.

But maybe I'm the only one who's bothered with code generation :D

My biggest bone to pick with this interview process (I was also interviewed, passed the interview and then trained to do interviews) is that the calibration is all over the place.

If you're lucky, you'll get an interviewer who thinks that if you can solve a case insensitive palindrome, it means you should be hired with strong confidence.

But if you're unlucky, you'll end up with a person who will squeeze the last ounce of blood out of you, and even if you solve 99% of the questions correctly, he will still think you're low confident or even a no-hire because you missed that one edge case...

I have no affiliation with the project. Just found it, tried it out, and it looks very promising...

Another article about the same thing: https://www.newyorker.com/tech/elements/the-whistled-languag...