I've shipped three software projects in the last two months. I shipped zero in the preceding year.
Right now I'm building a system to decompile MAME ROM games into idiomatic JavaScript. It completes one game in about 3 days - complete with extensive commenting. I started it about 2 weeks ago.
I'm not sure "EVERYTHING HAS CHANGED" but if you're not seeing dramatic change, I suspect you're not looking.
> if you're not seeing dramatic change, I suspect you're not looking
If the people claiming that everything has changed were even a fraction as productive as they think, then it would be visible even to people who weren't looking.
As it is, a lot of people are actively looking but still not seeing it.
I'd love for HN to explain why I was flagged here for answering:
> If the people claiming that everything has changed were even a fraction as productive as they think, then it would be visible even to people who weren't looking.
with:
> If you can't see this - you should probably get your eyes checked.
He is clearly suggesting I am delusional about my productivity gains, and I'm the one who gets flagged.
Don't be a jerk. A lot of people have had bad experiences with these systems from people basically saying exactly what you're saying. This "you should get your eyes checked" mentality is pervasive with the nerds who confidently say that their analysis is solid when it's... not. It's a big problem in communities that have non-conventional videogame puzzles for example. Someone will come in, announce some novel solution.... leading to a four hour fight about the efficacy of LLMs, only to find a mistake in their code the next day.. never to be seen again.
Mind you, I'm not saying that these systems aren't great at reverse engineering and whatnot. They're spectacularly good at that. But RE is a relatively constrained problem because everything is still right in front of you. For complicated problems where the signal is mixed between noise... less so.
My point is just that you're talking past people and being an asshole in the process. Not everything is about winning an argument. Lower your temperature and you'll have more productive discussions that revolve around disagreements.
I'm not disputing that it's a hard thing to do, but we had decompilers before; "miracle" is rather stretching it.
Even granting that framing though, the claimed increase in productivity isn't a one-off, but asserted for everyone using them; you're claiming mass-produced miracles but trying to depend on a single example.
I used to use dotPeek all the time back in like 2012. There are also things like Ghidra that the NSA made. I dunno, google them, there are a lot. I'm not saying LLMs aren't useful here, just that it's not a huge deal. Also, why are AI people always making tools to take other people's IP? You are kind of fucking with people's intellectual property without permission it seems like. I'm not saying it's the crime of the century or anything but it's distressing how many AI projects come out of "let me repackage something someone else made"
I'm seeing more examples of you saying you're providing evidence than of you actually providing evidence. So far I've heard a single anecdote about what you did.
That's a pretty far cry from the "evidence" you claim to have dropped on multiple comments.
The claim was this is not possible. I show it happening. That's a proof.
You're confused because the example is my own project. But I'm not merely describing it - I am giving you full access to it to confirm or reject my argument.
Not anecdotal. Saying it three times does not make it true.
What things have you tried that didn't go well? At this point, that's more interesting.
If you want to have a whack at a hard problem, have a try at the cryptographic puzzle in Noita if you want to see the silly failure modes of these systems. Lots of people have tried to solve it with AI and nothing's budged. If you can solve it, great.
If you'd like to change the subject I'd be happy to do so.
You're right - there are still many problems that have not been solved by AI. AI has not found a cheap and effective way to turn lead into gold, as an example.
But I sorta think that's a silly metric. There are always going to be unsolved problems for any system. The interesting metric is how many problems have been solved - and how surprising are they.
Like I said above - a decompiler that produces semantic understanding of the machine code it's looking at (this address is score; this address is how many lives are left; this address is...) would have been considered literally impossible before LLMs. Today, I was able to implement such a system in under two weeks.
That's an impressive result. If you disagree, I'd love to discuss it with you.
I'm aware of what they're good at. I got claude 4.6 to run linux as a native postgres module inside linux inside postgres inside linux inside postgres on a mac. It was impressive then and it's still impressive. IOW, I really think you're misinterpreting my comments as significantly more inflammatory than I actually mean. And you seem to be taking it personally.
But what I'm interested in is in the edges of their failure modes because those failure modes prop up frequently in pernicious ways. For a similar reason that I want to understand the edges of my own intellectual failure modes by regularly challenging myself. This conversation is an example of that.
It seems like you're just not interested in discussing the edges and failure modes. You call that a "silly metric", I call it, "understanding your tools".
My primary interest in these threads is to debunk the claims that "AI is a hoax".
Granted, that's not exactly the claim these days, the claim's goalposts shift constantly, as these things do.
But if you go back to the top of this conversation - I asked why people aren't seeing the miracles (not cute tricks - useful things that simply were impossible before) and I was largely met with the response "you're delusional if you see miracles."
But the thing is - I am not. So I am going to continue arguing that case every time I see a version of it.
So no, I'm not interested in exploring other issues. Sorry. Yours in particular I find as interesting as discussing whether Microsoft Word can be used to edit video files. Not interesting.
But if you need validation, here it is: there are many problems that AI cannot solve. Of course there are. There always will be.
HEH. Not sure why you're implying I have changed my position. I said as much above and would have said the same at any time. I don't think it's controversial in the least that there are things AI can't do.
Do you have a weird need to reframe this so you can feel like you won? You shouldn't do that. It's not healthy.
Oh hey - the timer finally kicked off. In case you didn't see my reply - here you go:
HEH. Not sure why you're implying I have changed my position. I said as much above and would have said the same at any time. I don't think it's controversial in the least that there are things AI can't do.
Do you have a weird need to reframe this so you can feel like you won? You shouldn't do that. It's not healthy.
What counter example did you provide? You said you had tried a few projects. You can just claim that without any evidence. I have no reason to believe you're not lying.
You could say "I used AI to cure cancer." This isn't a counter example, because there's no evidence for it.
I'm sorry downvoters - it's true. In his opening statement he complained about nastiness and then became extremely nasty himself, when I had said nothing like that to him. Other than just disagree and provide evidence.
But "OH MY GOD EVERYTHING HAS CHANGED THE OLD WAY IS DEAD 10X MORE PRODUCTIVE" - no. If that were true, we'd notice it in the software all around us.