Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Don't fall into the anti-AI hype

antirez.com

Post reply on HN
Print view View original

951–960 of 1001 posts

Re: Don't fall into the anti-AI hype

#951

Post by procflora »

themk wrote: post #536
ironman1478 wrote: post #348

I'm not sure what to make of these technologies. I read about people doing all these things with them and it sounds impressive. Then when I use it, it feels like the tool produces junior level code unless I babysit it, then it really can produce what I want. If I have to do all this babysitting, is it really saving me anything other than typing the code? It hasn't felt like it yet and if anything it's scary because I…

> and reading code is harder than writing it. This is the things thar gets me the most. Code review is _hard_. So hard that I'm convinced my colleagues don't do it and just slap "LGTM" on everything. We are trading "one writer, one reader" for "two readers", and it seems like a bad deal.

Yep, and I'll add: the first reader is the first maintainer. When that is turned over to an LLM agent the organization's leadership had better be prepared to entertain rewrites (reprompts?) of significant portions of LLM-generated code on a regular basis. The call of the rewrite isn't new of course, but it'll be far more alluring since LLMs are at their most "productive" and least destructive when working from a clean slate.

Re: Don't fall into the anti-AI hype

#952

Post by switchbak »

simonw wrote: post #630
vignesh37 wrote: post #627

Earlier quoted context omitted.

The biggest frustration with LLMs for me is people telling me I'm not promoting it in a good way. Just think about any product where they are selling a half baked product, and repeatedly telling the user you are not using it properly.

But that's not how most products work. If you buy a table saw and can't figure out how to cut a straight line in a piece of wood with it - or keep cutting your fingers off - but didn't take any time at all to learn how to use it, that's on you. Likewise a car, you have to take lessons and a test before you can use those! Why should LLMs be any different?

It's not anyone's job to "promote it in a good way", we have no responsibility either for or against such tech.

The analogy would be more like: "yeah, the motor blew up and burned your garage, but please don't be negative - we need you to promote this saw in a good way".

Sure, it's important to "hold it right", but we're not in some cult here where we need to all sell this tech well beyond its current or future potential.

Re: Don't fall into the anti-AI hype

#953

Post by gspr »

Cthulhu_ wrote: post #828
gspr wrote: post #767

Earlier quoted context omitted.

I still really, really, really struggle to see how humans are going to maintain and validate the programs written by LLMs if we no longer know (intimately) how to program. Any thoughts?

Same how we do it now - look at the end result, test it. Testers never went away. Besides, your comment goes by the assumption that we no longer know (intimately) how to program - is that true? I don't know C or assembly or whatever very well, but I'm still a valuable worker because I know other things. I mean it could be partially true - but it's like having years of access to Google to quickly find just what I need…

> Besides, your comment goes by the assumption that we no longer know (intimately) how to program - is that true? I don't know C or assembly or whatever very well, but I'm still a valuable worker because I know other things.

The proposal seems to be for LLMs to take over the task of coding. I posit that if you do not code, you will not gain the skills to do so well.

> I mean it could be partially true - but it's like having years of access to Google to quickly find just what I need, meaning I never learned how to read e.g. books on software development or scientific paper end to end.

I think you've misunderstood what papers are for or what "the previous generation" used them for. It is certainly possible to extract something useful from a paper without understanding what's going on. Googling can certainly help you. That's good. And useful. But not the main point of the paper.

Re: Don't fall into the anti-AI hype

#954

Post by alexsmirnov »

totallykvothe wrote: post #569

I don't understand the stance that AI currently is able to automate away non-trivial coding tasks. I've tried this consistently since GPT 3.5 came out, with every single SOTA model up to GPT 5.1 Codex Max and Opus 4.5. Every single time, I get something that works, yes, but then when I start self-reviewing the code, preparing to submit it to coworkers, I end up rewriting about 70% of the thing. So many important deta…

This is exact the impression that I got. Every question or task given to LLM returns pretty reasonable, but flawed result. For the coding, those are hard to spot but dangerous mistakes. They all look good and perfectly reasonable, but just wrong. Anthropic compared Claude Code to a "slot machine", and I fell that AI coding now is something close to gambling addiction. As small wins keep gambler to make more bets, so correct results from AI keep developers to use it: "I see it made correct solution, let's try again!" At a startup CTO, I review most of the pull requests from team members, and team uses AI tools actively. The overall picture strongly confirms your second conclusion.

Re: Don't fall into the anti-AI hype

#955

Post by simonw »

alexsmirnov wrote: post #954
totallykvothe wrote: post #569

I don't understand the stance that AI currently is able to automate away non-trivial coding tasks. I've tried this consistently since GPT 3.5 came out, with every single SOTA model up to GPT 5.1 Codex Max and Opus 4.5. Every single time, I get something that works, yes, but then when I start self-reviewing the code, preparing to submit it to coworkers, I end up rewriting about 70% of the thing. So many important deta…

This is exact the impression that I got. Every question or task given to LLM returns pretty reasonable, but flawed result. For the coding, those are hard to spot but dangerous mistakes. They all look good and perfectly reasonable, but just wrong. Anthropic compared Claude Code to a "slot machine", and I fell that AI coding now is something close to gambling addiction. As small wins keep gambler to make more bets, so…

If someone gives you access to a slot machine which is weighted such that it pays out way more than you put into it, my advice is to start cranking that lever.

If it does indeed start costing more than it's paying out, step away.

Re: Don't fall into the anti-AI hype

#956

Post by throw1235435 »

echelon wrote: post #343

I love Antirez. > However, this technology is far too important to be in the hands of a few companies. This is the most important assessment and we should all heed this warning with great care. If we think hyperscalers are bad, imagine what happens if they control and dictate the entire future. Our cellphones are prisons. We have no fundamental control, and we can't freely distribute software amongst ourselves. Every…

This is what will occur - the bad scenario that is. Labor and its knowledge distributes (hard to contain knowledge), capital centralises and compounds. Always been that way. With AI there will be a a tension between the two of course.

The root question is: Will AI decentralise quicker than the disruption to this profession? I don't think so.

I've noticed us techies don't really understand economics and game theory all that well - we just see awesome toy and want to play with it and want others to enjoy it too. We have worked to democratize computing for years (e.g. OSS) now to our detriment. No one in society long term respects people who do this in a capitalist system; they find them naive. I can now understand why other professions find us a little immature like kids playing with tech toys.

I love solving problems with technology and love the field, but as I've gotten older I look back on a less technological life with nostalgia. Technology for all its benefit has disrupted the one thing humans do need and had for millions of years in our evolution - relative stability within their lifetimes. The mental health benefits to stability are massive and usually unmeasured. Technology, as evidenced by this thread, creates more and more anxiety about our future and our place within the community (e.g. social media, AI, and others). "Adaptability" isn't just a psychological trait; a wealthy person and secure person by definition is more adaptable too.

Re: Don't fall into the anti-AI hype

#957

Post by kahrl »

systemf_omega wrote: post #8

What I don't understand about this whole "get on board the AI train or get left behind" narrative, what advantage does an early adopter have for AI tools? The way I see it, I can just start using AI once they get good enough for my type of work. Until then I'm continuing to learn instead of letting my brain atrophy.

When I think real hard about it, I can translate the bullshit speak. "get on board the AI train or get left behind" -> "BUY BUY BUY OUR SHIT! WE ARE SOOO OVER LEVERAGED WE ARE SO FUCKED. PLEASE CONSOOM"

Re: Don't fall into the anti-AI hype

#958

Post by DamnInteresting »

simonw wrote: post #938
DamnInteresting wrote: post #911

Earlier quoted context omitted.

This has been my suspicion since LLMs began eating the Internet. Whether it's code or writing, now that LLMs are consuming their own output, the Habsburg Jaw[1] is going to quickly become evident. It is very difficult--sometimes impossible--to know whether a given chunk of input is wholly or partially generated by an LLM. Nevertheless, filtering input may become a critical task. That expense will be passed to the con…

Nothing I've seen from the AI labs appears to indicate that they are worried about model collapse in the slightest. That makes sense to me, because if their models start getting worse because there's slop in the training data they can detect that and take steps to fix it. Their entire research pipeline is about finding what makes models that score better! Why would they keep going with a technique that scored worse?

> Nothing I've seen from the AI labs appears to indicate that they are worried about model collapse in the slightest.

AI labs are insufferable hype machines, they are unlikely to sow doubt about their own business models.

> they can detect that and take steps to fix it.

Each model will need an endless diet of new content to remain relevant, and over time, avoiding ingestion of LLM output (and the accompanying inbreeding depression) will likely be a tricky proposition. Not impossible, but expensive and error-prone.

Re: Don't fall into the anti-AI hype

#959

Post by upcoming-sesame »

mvkel wrote: post #580
totallykvothe wrote: post #569

I don't understand the stance that AI currently is able to automate away non-trivial coding tasks. I've tried this consistently since GPT 3.5 came out, with every single SOTA model up to GPT 5.1 Codex Max and Opus 4.5. Every single time, I get something that works, yes, but then when I start self-reviewing the code, preparing to submit it to coworkers, I end up rewriting about 70% of the thing. So many important deta…

> Other people are just less picky than I am I think this is part of it. When coding style has been established among a team, or within an app, there are a lot of extra hoops to jump through, just to get it to look The Right Way, with no detectable benefit to the user. If you put those choices aside and simply say: does it accomplish the goal per the spec (and is safe and scalable[0]), then you can get away with a lo…

you don't even have to put these choices aside too much, you can have very detailed linting rules that nudge the LLM towards the style you want.

Re: Don't fall into the anti-AI hype

#960

Post by mvkel »

saltyoutburst wrote: post #669
mvkel wrote: post #580

Earlier quoted context omitted.

> Other people are just less picky than I am I think this is part of it. When coding style has been established among a team, or within an app, there are a lot of extra hoops to jump through, just to get it to look The Right Way, with no detectable benefit to the user. If you put those choices aside and simply say: does it accomplish the goal per the spec (and is safe and scalable[0]), then you can get away with a lo…

Do you have any references for "vibe coded monoliths tend to collapse in on themselves at ~30,000 LOC"? I haven't personally vibed up anything with that many LOC, so I'm legitimately curious if we have solid numbers yet for when this starts to happen (and for which definitions of "collapse").

Just my experience in vibe coding apps from gpt-3.5 onwards (mostly NextJS or Node). In gpt-3.5, I had to really hand-hold it, getting it to write one function at a time, then a separate task to glue the functions together.

Now, it can build almost all of an app from a single prompt, but will start to rewrite utility functions, or modules, forgetting that they already exist. Some of this is still solvable with clever prompting, but if you're just attacking it without thinking, ~30,000 LOC seems to be the app 'size' that it will start to exhibit those behaviors.

Post reply on HN