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

671–680 of 1001 posts

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

#671

Post by trueno »

Aeolun wrote: post #644
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…

Maybe if your coding style is already close to what an LLM like Claude outputs, you’ll never have these issues? At least it generally seems to be doing what I would do myself. Most of the architectural failures come from it still not having the whole codebase in mind when changing stuff.

I actually think it's less about code style and more about the disjointed way end outcomes seem to be the culmination of a lot of prompt attempts over the course of a project/implementation.

The funny thing is reviewing stuff claude has made isn't actually unfamiliar to me in the slightest. It's something I'm intimately familiar with and have been intimately familiar with for many years, long before this AI stuff blew up...

..it's what code I've reviewed/maintained/rejected looks like when a consulting company was brought on board to build something. Such a company that leverages probably underpaid and overworked laborers both overseas and US based workers on visas. The delivered documentation/code is noisy+disjointed.

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

#672

Post by furyofantares »

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…

I think you should try harder to find their limits. Be as picky as you want, but don't just take over after it gave you something you didn't like. Try again with a prompt that talks about the parts you think were bad the first time. I don't mean iterate with it, I mean start over with a brand new prompt. Try to figure out if there is a prompt that would have given you the result you wanted from the start.

It won't be worth it the first few times you try this, and you may not get it to where you want it. I think you might be pickier than others and you might be giving it harder problems, but I also bet you could get better results out of the box after you do this with a few problems.

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

#673

Post by zephyrthenoble »

It was weird to read this. I know antirez is on HN, so it's strange to say this, but here goes...

I always looked up to antirez. Redis was really taking off after I graduated and I was impressed by the whole system and the person behind it. I was impressed to see them walk away to do something different after being so successful. I was impressed to read their blog about tackling difficult problems and how they solved them.

I'm not a 10x programmer. I don't chase MVPs or shipping features. I like when my manager isn't paying attention and I can dig into a problem and just try things out. Our database queries have issues? Maybe I can write my own AST by parsing just part of the code. Things like that.

I love BUILDING, not SHIPPING. I learn and grow when I code. Maybe my job will require me to vibe code everything some day just to keep up with the juniors, but in my free time I will use AI only enough to help speed up my typing. Every vibe coded app I've made has been unmaintainable spaghetti and it takes the joy out of it. What's the point of that?

To bring it all together, I guess some part of me was disappointed to see a person that I considered a really good programmer, seem to indicate that they didn't care about doing the actual programming?

> Writing code is no longer needed for the most part

> As a programmer, I want to write more open source than ever, now.

This is the mentality of the big companies pushing AI. Write more code faster. Make more things faster. Get paid the same, understand less, get woken up in the middle of the night when your brittle AI code breaks.

Maybe that's why antirez is so prolific and I'm not.

Sometimes I wish I was a computer scientist, instead of a programmer...

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

#674

Post by p0w3n3d » 

  > How do I feel, about all the code I wrote that was ingested by LLMs? I feel great to be part of that, because I see this as a continuation of what I tried to do all my life: democratizing code, systems, knowledge.
I don't see it as democratic or democratising. TBH the knowledge is stored in three giga companies that used sometimes almost non-lawful (if not lawful?) methods to gain it, scraping it off the gpl projects etc. And now they are selling it to us without giving the models away. The cost IS understandable because the horrendously expensive vector cards do not come for free, but there is only one country the knowledge is gathered in so this might as well fade away one day when an orange present says so (gimme all the monies or else..)

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

#675

Post by troupo »

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?

Table saws and cars are deterministic. Once uou learn how to use them, the experience is repeatable.

The various magic incantations that LLMs require cannot be learned or repeated. Whatever the "just one more prompt bro" du jour you're thinking of may or may not work at any given time for any given project in any given language.

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

#676

Post by never_inline »

irthomasthomas wrote: post #366
embedding-shape wrote: post #95

> But what was the fire inside you, when you coded till night to see your project working? It was building. I feel like this is not the same for everyone. For some people, the "fire" is literally about "I control a computer", for others "I'm solving a problem for others", and yet for others "I made something that made others smile/cry/feel emotions" and so on. I think there is a section of programmer who actually do…

In my feed 'AI hype' outnumbers 'anti-AI hype' 5-1. And anti-hype moderates like antirez and simonw are rare. To be a radical in ai is to believe that ai tools offer a modest but growing net positive utility to a modest but growing subset of hackers and professionals

The only AI bloggers who don't have something to sell seems to be simonw, the flask guy, and this redis guy. Any other blog recommendations from HN?

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

#677

Post by cookiengineer »

zephyrthenoble wrote: post #673

It was weird to read this. I know antirez is on HN, so it's strange to say this, but here goes... I always looked up to antirez. Redis was really taking off after I graduated and I was impressed by the whole system and the person behind it. I was impressed to see them walk away to do something different after being so successful. I was impressed to read their blog about tackling difficult problems and how they solved…

> To bring it all together, I guess some part of me was disappointed to see a person that I considered a really good programmer, seem to indicate that they didn't care about doing the actual programming?

My take on this is that we as a society are now on the verge of transitioning towards programming as an art form. And the methodologies of art vs non art programming are vastly different.

Take clothes, for example. Manufacturing is vastly optimized for throughput, but its art form is heavily optimized for design and customization. Maybe that is what all this is about now with programming, too?

I too would think of myself as someone who likes to code for the sake of explorative understanding and optimization. I'm pretty bad at the last 10%, like _reeeally_ bad actually.

But I am aware that the methodology of programming is changing. And currently I believe that design and customization might in parts also change, because a lot of LLM- / slop-coded successful projects were optimizing for something like text-in-the-loop where they started with a terminal CLI and made it a real design later, because the LLM agent was able to parse and understand CLI / TTY characters.

Maybe this is what it's actually about. Maybe we need to optimize things for text now so that LLMs can help us more in these topics?

I'm thinking lately a lot about scene graphs and event graphs and how to make them serializable so that I can be more efficient in generating UIs. Sorry for babbling, maybe these are just thoughts I'm gonna regret in the future.

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

#679

Post by redox99 »

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…

It sounds harsh but you're most likely using it wrong.

1) Have an AGENTS.md that describes not just the project structure, but also the product and business (what does it do, who is it for, etc). People expect LLMs to read a snippet of code and be as good as an employee who has implicit understanding of the whole business. You must give it all that information. Tell it to use good practices (DRY, KISS, etc). Add patterns it should use or avoid as you go.

2) It must have source access to anything it interacts with. Use Monorepo, Workspaces, etc.

3) Most important of all, everything must be setup so the agent can iterate, test and validate it's changes. It will make mistakes all the time, just like a human does (even basic syntax errors), but it will iterate and end up on a good solution. It's incorrect to assume it will make perfect code blindly without building, linting, testing, and iterating on it. No human would either. The LLM should be able to determine if a task was completed successfully or not.

4) It is not expected to always one shot perfect code. If you value quality, you will glance at it, and sometimes ahve to reply to make it this other way, extract this, refactor that. Having said that, you shouldn't need to write a single line of code (I haven't for months).

Using LLMs correctly allow you to complete tasks in minutes that would take hours, days, or even weeks, with higher quality and less errors.

Use Opus 4.5 with other LLMs as a fallback when Opus is being dumb.

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

#680

Post by smj-edison »

onlyrealcuzzo wrote: post #591
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…

I'm not sure if I got in this weird LLM bubble where they give me bad advice to drive engagement, because I can't resist trying to correct them and tell them how absurdly wrong they are. But it is astounding how terrible they are at debugging non-trivial assembly in my experience. Anyone else have input here? Am I in a weird bubble? Or is this just not their forte? It's truly incredible how thoughtless they can be, s…

I've tried to use Claude Code with Sonnet 4.5 for implementing a new interpreter, and man is it bad with reference counting. Granted, I'm doing it in Zig, so there's not as much training, but Claude will suggest the most stupid changes. All it does is make the rare case of incorrect reference counting more rare, not fixing the underlying problem. It kept heaping on more and more hacks, until I decided enough is enough and rolled up my sleeves. I still can't tell if it makes me faster, or if I'm faster.

Even when refactoring, it would change all my comments, which is really annoying, as I put a lot of thought into my comments. Plus, the time it took to do each refactoring step was about how long it would take me, and when I do it I get the additional benefit of feeling when I'm repeating code too often.

So, I'm not using it for now, except for isolating bugs. It's addicting having it work on it for me, but I end up feeling disconnected and then something inevitably goes wrong.

Post reply on HN