FizzyOrange @ FizzyOrange @programming.dev

Posts

1

Comments

464

Joined

2 yr. ago

Oh you mean when I said this?

I expect it helps people of all experience levels fairly equally, but only with tasks that are relatively simple.

No I don't have actual data, just direct personal experience of asking AI to do simple and complex tasks - it does much better on simple tasks, especially in very widely discussed domains (HTML, CSS, JavaScript, Python etc.) Ask it any SystemVerilog stuff and it gets it wrong almost every time annoyingly!

What do you mean? I've seen people say that all the time on HN. No I'm not going to go and search for comments.

Yeah there are pros and cons. Desktop apps are not sandboxed. Mobile apps are often missing features and are annoying to install. Websites often have poor performance or janky UX on mobile, and you need to be online, and you don't have control of their availability.

I think the best option depends on what the thing is - ordering food from a random pub? Web site. Video editing? App.

Proper reasoning is always needed if you want a guarantee.

You formally verify your regexes? Doubtful.

Does that imply anything at all in LLMs’ favour?

Yes it suggest lower cognitive load.

Ok I can see you haven't actually come across any complex regexes yet...

(Which is probably a good thing tbh - if you're writing complex regexes you're doing it wrong.)

This is stupid pedantry. By that logic literally nothing is complex because everything is made up of simple parts.

Damn there are so many AI critics who have clearly not seriously tried it. It's like the smartphone naysayers of 2007 but much much worse.

You don't have to. You can read it.

Regexes aren’t hard to write, their logic is quite simple.

He did say complex regex. A complex regex is not simple.

Why? That is a great use for AI. I'm guessing you are imagining that people are just blindly asking for unit tests and not even reading the results? Obviously don't do that.

I don't think that's true. In fact most people say the opposite - AI doesn't help junior devs because they can't recognise when it's bullshitting. I don't really believe that either - that's just ego talking. I expect it helps people of all experience levels fairly equally, but only with tasks that are relatively simple. It's not like senior engineers never do those though.

It definitely is... But it's possible to be the most rigorous study and also not really prove anything. Proving this sort of stuff is ridiculously hard and expensive. We don't have proofs for even the most obvious things in programming, like that comments and good variable naming help comprehension. Sometimes studies even find the opposite.

Google tried to do that with Dart, and failed. In fairness Dart 1 was much worse than Dart 2... So maybe that was a good thing because there's no way they'd have been able to improve Dart as much as they have if it was part of the web.

For dates there finally is something better anyway: https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Temporal

Ha this is even worse than I could have imagined!

Yeah... This doesn't sound promising. Refusing to acknowledge the obvious fact that AI and hostile mods have driven away 99% (not an exaggeration!) of their audience. A visual makeover before the change anything. Trying to sell the Q/A database for AI despite the fact that you can download it for free. They even talk up their job advertising product that they inexplicably cancelled a few years ago (btw I found levels.fyi has a pretty good job database if anyone is looking).

If it were me I would:

1. Use AI to improve question quality - if people post obviously bad questions get AI to improve it via a conversation.
2. Make it waaaaay harder to close questions. Like, require 10 votes and allow the asker to reopen it for free once.
3. Make it so questions can't go below 0 points.
4. Make it impossible to close as duplicate. You should be able to mark questions as possible duplicates (and the asker can say "yes it is") but if they don't it should stay open.
5. Maybe even make it impossible to close questions at all. What actual purpose does it solve apart from driving people away?

Apart from the AI that's all stuff they should have done 10 years ago.

It's probably too late anyway.

You shouldn’t be pulling an external project as a submodule, that’s just coupling yourself way way too tightly to external code.

You're no more tightly coupled than if you zip that repo up, and put it on an internal server. It's the exact same code you've just changed the distribution method.

And my whole point is that wouldn't be necessary if Git had a version of submodules that worked properly!

You guys seriously lack imagination.

Not my experience. More like 50% coding/debugging, 25% meetings, 25% admin/helping people/procrastinating.

Yes I'm aware where Git came from.

Large files don’t work with git, as it always stores the whole history on your drive.

Not any more. That's only the default. Git supports sparse checkouts and blobless checkouts both of which only get a subset of a repo. And actually it has supported --depth for as long as I remember.

Libraries are not always a suitable solution. You just haven't worked on the same projects I have and you can't imagine all the things submodules are used for.

On top of that, I can't force all third party projects to turn their repos into nice easily installable packages. Especially if they're using a language that doesn't have a package manager.