Even if LLMs write more buggy code they can still bring up software quality in the short to medium term by allowing you to clear out a lot of the backlog of bugs and UI issues that are known but never had enough priority to be fixed
Debugging and developing first fixes is also one of the spaces where current LLMs are the biggest force multipliers. Especially if you have reproduction cases the LLM can test on its own
But long-term it might look very different as more and more of the code becomes LLM written
Make sense to me. I can see how LLMs can help you make better systems. I don't have a christal ball but I can see how focusing on speed (or more precisely volume) can have a lot of unintended consequences.
CCC was and is a marketing stunt for a new model launch. Impressive, but still suffers from the same 80:20 rule. These 20% are optimizations, and we all know where the devel in “let me write my own language”.
I read the whole thing, but I have a rule: I stop reading any review that includes “huge fan of <any_software>.”
Like many things in life, getting too emotional isn’t usually helpful. :)
I really do appreciate that younger generations are eager to write new software and push things forward — but reinventing lukewarm water isn’t the way to go. And yet, here we are, deep into the era of exactly that.
That said, it seems like Kitty and tmux are actually quite different tools after all.
It does feels like a good use of AI (ML, really) would be to write a "disaggregator" for HN that tags submissions by category and lets users browse the bits they care about. Wish I had time to do it....
Thanks for the reply! The extension seems great at first but it doesn't let me filter out tags and basically just redirects me to your domain, so it's not really how I expect an extension to behave.
And a TL;DR, both for the articles themselves as well as the discussions.
I don't think a TL;DR can replace most articles that appear on HN, but it can certainly tell me whether the article is interesting, much better than any headline ever could. Especially so if the TL;DR is written by a neutral AI with no interest in making me click anything, and hence no qualms about surfacing the most important information to the top.
I actually tried to do this, but it was with GPT-3.5, and I didn't exactly like how it worked. I should look at this again, I wouldn't be surprised if the code I used back then could just be ported over to 2.5 Flash and produce much better results.
Because the language of the week changes often, and learning can be done by solving the problems of today instead of rewriting software into a version that will never be used. I mean... who still uses all the rewrites to ruby?
Even emacs was rewritten to rust ( https://github.com/remacs/remacs ), many hours were spent, and the last actual code commit was 5 years ago.... why not spend that time by making the "normal" emacs better? Or make something new in rust?
Well, it's actually just a hardcoded slideshow of E1M1 while something vaguely approximating the main riff of At Doom's Gate plays inconsistently in the background, but you'll have to watch all 15 excruciating minutes of this poorly-narrated Youtube video I'm linking to figure that out.
I ported DOOM to it. In 100 LOC. BTW it's just a a line shooting a ball of zero width, at another line. And there's some movement left and right. But not forward, nor backwards. So there's no real strafing. And the other line doesn't shoot a ball of zero width back.
I started reading hackernews from very old posts to new ones, so i'm still rewriting stuff to ruby, because that will definitely be the universal programming language for the future!
That's how you start. It's also important to recognize that much of the UNIX and eventually POSIX framework was laid out. They were filling in a small gap and making it free. Which is really the most amazing part to me.
reply