Those things didn’t destroy communities, pollute the earth, wrestle personal computing away from the populace, use up all the drinking water in an area, and provide a near total and realtime panopticon of everyone, everywhere, at all times, while stealing all the collected works of said society in order to be built without penalty at a time when ordinary folks are ordered to pay hundreds of thousands of dollars because they posted a social media video of their kid dancing to a song that was playing on broadcast radio.

But sure keep boiling in that pot because you don’t need to do all the boilerplate for your fucking Node project or whatever. Fucking frog.

None of what you brought up as a positive are things an LLM does. Most of those things existed before the modern transformer-based LLMs were even a thing.

LLM-s are glorified text prediction engines and nothing about their nature makes them excel at formal languages. It doesn't know any rules. It doesn't have any internal logic. For example if the training data consistently exhibits the same flawed piece of code then an LLM will spit out the same flawed piece of code, because that's the most likely continuation of its current "train of thought". You would have to fine-tune the model around all those flaws and then hope some combination of a prompt won't lead the model back into that flawed data.

I've used LLMs to generate SQL, which according to you is something they should excel at, and I've had to fix literal syntax errors that would prevent the statement from executing. A regular SQL linter would instantly pick up that the SQL is wrong but an LLM can't pick up those errors because an LLM does not understand the syntax.

The referenced job is clearly talking about the current over valued tech bro kind, you buffoon.

Neither does a locally run LLM model.

We've got studies that show AI makes you feel more productive while you're actually less productive. And all you're offering is a feeling you feel. Get high on your own supply if you want, but don't drag down good companies with your evangelism.

Honestly if part of their job is at all trying to get old shit to run on new operating systems AI is very useful for that task.

Part of my job is keeping a 30 year old c++ application compiling and building on newer versions of Linux. LLMs have made this a far easier experience.

There you go, predictably making more baseless claims. If these things are supposedly so great, prove it.

And how did you hallucinate a misspelling in my comment? Maybe take a break from Elon's CSAM bot for a while.

I'm curious about these studies. Do you have a citation?

Job Requirements:

Active use of AI tools in daily development workflows, and enthusiasm for helping the team increase adoption

Nice to have:

Passion for games and game preservation

AI Mandatory, game preservation optional. Glad they got their priorities straight

I don't want to say you're totally wrong, but I am skeptical of the benefit. Sure, maybe it works now, which is cool, but is it making changes that are maintainable? The next time someone does this is it going to work? If we just constantly have LLMs update code, when does it start breaking, and when it does is it going to be in a state someone can fix?

Hey Steven, how do you think they make those models?

(As if you genuinely believe those are the ones GOG is using.)

Im not generally making source code changes. It's the dependencies.

Mainly we're talking about building very old versions of things like libpng. Making things like autoconf and configure and cmake all work is a pain in the ass as their versions slowly change.

The business would be content to let it run on Ubuntu 12 until it's a major problem so I can't let the perfect be the enemy of good.

Copilot business subscriptions have fairly granular usage tracking, so they'd probably just replace you right away with someone who isn't quite so reserved. Looking at the comments here and in other places, there is certainly no shortage of such people.