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

The cult of vibe coding is dogfooding run amok

bramcohen.com

Post reply on HN
Print view View original

541–550 of 555 posts

Re: The cult of vibe coding is dogfooding run amok

#541

Post by nephanth »

tombert wrote: post #269
xienze wrote: post #258

Earlier quoted context omitted.

> I suspect if people saw the handwritten code of many, many, many products that they used every day they would be shocked. Absolutely. The difference is that the amount of bad code that could be generated had an upper limit on it — how fast a human can type it out. With LLMs bad code can be shat out at warp speed.

Oh I don't disagree with that. I am getting pretty tired of people making multi-thousand-line pull requests with lots of clearly AI-generated code and expecting it to be merged in. I think the better unit to commit and work with is the prompt itself, and I think that the prompt is the thing that should be PR'd at this point, because ultimately the spec is what's important.

But then who's responsible for reviewing and correcting the ai-written code ?

Re: The cult of vibe coding is dogfooding run amok

#542

Post by nephanth »

thesz wrote: post #506
tombert wrote: post #297

Earlier quoted context omitted.

If you make the prompts specific enough and provide tests that it has to run before it passes, then it should be fairly close to deterministic. Also, people aren't actually reading through most of the code that is generated or merged, so if there's a fear of deploying buggy code generated by AI, then I assure you that's already happening. A lot.

> If you make the prompts specific enough and provide tests that it has to run before it passes, then it should be fairly close to deterministic. Your prompt may work on the specific state of code base and not before or after some changes. Your tests can check for the specific behavior but not for the absence of undesirable behaviors induced by absence of some specific code or by addition of other specific code. > I…

Hell sometimes the ai may look at values in .env or other uncommited stuff. Its execution path may depend on which commands are available in the environment. The specific model and settings. On top of the inherent randomness if you havent set model heat to 0. i suspect it is very hard to get consistent reproducible ai runs

Re: The cult of vibe coding is dogfooding run amok

#543

Post by mattmanser »

bearjaws wrote: post #389
mattmanser wrote: post #162

Earlier quoted context omitted.

That's not actually true. When you move to the enterprise layer, suddenly you get the opposite problem, you have a low amount of "users" but you often need a load of CPU intensive or DB intensive processing to happen quickly. One company I worked for had their system built by, ummmm, not the greatest engineers and were literally running out of time in the day to run their program. Every client was scheduled over 24 h…

Without seeing more this seems like it could be solved by not recomputing the entire history to add on data. Depends what kind of math you are doing however. Some sort of check point system could likely save significant IO. What am I missing that requires you to recompute all data every day?

I can't go into too much detail because of NDA.

It was receiving huge volumes of data from each financial client, and matching it all up to try and find certain things. And match it with existing historical data. Not ads or online tracking. So the loop was adding this data and recalculating everything, It had to be done sequentially, I can't remember the exact reason, but it was a good one.

I was only there a few months as they were so dysfunctional I jumped to another job offer I'd received. We were having endless sprint meetings to "plan" this work when all it needed was someone experienced like me refactoring it for a coupe of days. There was a lot of junior devs with senior developer titles as everyone invariably got promoted every year. The funny thing about the sprint cards was all the tasks I put up a 1 for, the other developers put up a 10, and all the ones I put up 10s for the other developers put up 1s. That's what happens when you let junior devs have a say, no comprehension of what's hard or not.

Before I went I did point out the multiple pretty obvious O(n²) loops they had in the main calculation loop the results of which could easily be cached but I don't know if it went whoosh over their heads.

I'm pretty certain if they'd just let me got on with it instead of holding up sprint cards in the meetings they'd have been down to doing the whole lot is a 1/2 hour run a day, even in the short time I was there.

In my experience the first run of optimizing something like that usually doesn't take long and has huge benefits.

Re: The cult of vibe coding is dogfooding run amok

#544

Post by deterministic »

marcus_holmes wrote: post #438
tombert wrote: post #88

Earlier quoted context omitted.

I suspect if people saw the handwritten code of many, many, many products that they used every day they would be shocked. I've worked at BigCos and startups, and a lot of the terrible code that makes it to production was shocking when I first started. This isn't a dig at anyone, I've certainly shipped my share of bad code as well. Deadlines, despite my wishes sometimes, continue to exist. Sometimes you have to ship a…

It's an unpopular truth for our industry, but the point of commercial software development is not to write good code; it's to write profitable code. There are some cases where the most profitable code is also good code. We like those. But in most (99%+) cases, the code is not going to survive contact with the market and so spending any time on making it good is wasted.

In my experience (30+ years) writing good code (and good automatic tests) makes me way more productive.

Both when initially writing the code and later when maintaining it.

Good code dramatically reduce bugs and makes the remaining bugs more visible.

I spend almost zero time fixing bugs. Because there aren't many. Not a brag. Just the truth.

Re: The cult of vibe coding is dogfooding run amok

#545

Post by glasswerkskimny »

hintymad wrote: post #107

It looks vibe coding, or at AI coding in general, has been challenging a few empirical laws: - Brooks' No Silver Bullet: no single technology or management technique will yield a 10-fold productivity improvement in software development within a decade. If we write a spec that details everything we want, we would write soemthing as specific as code. Currently people seem to believe that a lot of the fundamentals are w…

[dead]

Re: The cult of vibe coding is dogfooding run amok

#546

Post by i_love_retros »

gitaarik wrote: post #535
i_love_retros wrote: post #497

Earlier quoted context omitted.

But before AI it was possible to keep things in check because pull requests were manageable in both size and frequency. Now we have people passing figma screen shots to copilot and Claude and putting up single shot pull requests that have what used to be 5 jira stories all in one. It's impossible to keep any semblance of consistency and correctness in the codebase. Its also exhausting for the people that actually car…

The Pull Requests were maybe manageable, but the backlog wasn't ;). AI does help with that. And the tools that require more stability you can still spend more time on and review the code.

I'm guessing you're a manager and don't write much code or take interest in software quality

Re: The cult of vibe coding is dogfooding run amok

#547

Post by i_love_retros »

bakugo wrote: post #482
i_love_retros wrote: post #364

Earlier quoted context omitted.

I think you're all fucking crazy. Or maybe I am? I can literally see my teams codebase becoming an unmaintainable nightmare in front of my eyes each day. I use copilot and Claude code and I frequently have to throw away their massively verbose and ridiculously complex code and engage my withering brain to come up with the correct solution that is 80% less code. I probably get to the solution in the same time when all…

> What are we doing here? Trying to get rich quick. That's all this really is. When your goal is to just make as much money as possible in the shortest amount of time, code quality doesn't matter because not only will your product not be relevant long enough for maintainability to become a problem, your users likely aren't using it because it's good anyway, they're using it because of the marketing and hype around it…

AI psychosis is real. My whole team has it

Re: The cult of vibe coding is dogfooding run amok

#549

Post by bccdee »

tedmiston wrote: post #328
bccdee wrote: post #237

Earlier quoted context omitted.

One truism about coding agents is that they struggle to work with bad code. Code quality matters as much as always, the experts say, and AI agents (left unfettered) produce bad code at an unprecedented rate. That's why good practices matter so much! If you use specs and test it like so and blah blah blah, that makes it all sustainable. And if anyone knows how to do it right, presumably it's Anthropic. This codebase h…

i think you are conflating anthropic (the startup) with claude code (the leaked source of one of said startup's products) i.e., the claude code codebase doesn't need to be good right now [^1] — so i don't think the assumption that this is an exemplary product / artifact of expert agentic coding actually holds up here specifically [^1]: the startup graveyard is full of dead startups with good code

I'm not taking a "zero technical debt" stance. Every successful startup eventually ends up treating their 1.0 codebase as a legacy burden. That's the cost of doing business, but it is a real cost. If Anthropic could have kept their code quality high with a few clever tricks and a few extra tokens, surely they would have.

So I take this whole episode as confirmation that code which is subject to high-velocity low-supervision work by agents decays quickly even under ideal circumstances.

Re: The cult of vibe coding is dogfooding run amok

#550

Post by cagatayk »

semicolon_storm wrote: post #14

It’s truly strange that people keep citing the quality of Claude code’s leaked source as if it’s proof vibe coding doesn’t work. If anything, it’s the exact opposite. It shows that you can build a crazy popular & successful product while violating all the traditional rules about “good” code.

Code agents built by tech companies have very different cost functions than the users of those tools. For engineers working on Claude Code, tokens are practically free, unlike you and me. They probably count on tackling increasing complexity of a grotesquely large code base with better models.

It ends up being a race, where you build something ugly that makes money directly (subscriptions) and indirectly (hype -> investors throwing money at you), you use that money to hire people, who build scarier/better models, which you then use (and this is the new thing) to directly improve the ugly thing more, hoping to make it less ugly and better at making money faster than you are spending it.

It sounds exhausting to me, but it probably feels exhilarating to the people in the middle of it.

Post reply on HN