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

Generative AI coding tools and agents do not work for me

blog.miguelgrinberg.com

Post reply on HN
Print view View original

431–440 of 464 posts

Re: Generative AI coding tools and agents do not work for me

#431

Post by dvt »

skydhash wrote: post #309
dvt wrote: post #64

I'm actually quite bearish on AI in the generative space, but even I have to admit that writing boilerplate is "N" times faster using AI (use your favorite N). I hate when people claim this without any proof, so literally today this is what I asked ChatGPT: write a stub for a react context based on this section (which will function as a modal): ``` // a bunch of stuff ``` Worked great, it created a few files (the hoo…

For this case, i will probably lift off the example from the library docs. Or spend 5 minutes writing a bare implementation as it would be all I need at the time. That’s an issue I have with generated code. More often, I start with a basic design that evolves based on the project needs. It’s an iterative process that can span the whole timeline. But with generated code, it’s a whole solution that fits the current nee…

> For this case, i will probably lift off the example from the library docs. Or spend 5 minutes writing a bare implementation as it would be all I need at the time.

Both of these would take longer than 5 minutes. There's also no "lifting" as this case involves both Provider and Context, so you'd have to combine React doc examples.

The only alternative would be knowing it by heart, which you might, but I don't (nor do I particularly care to). There's definitely a force multiplier here, even if just in the boring boilerplate cases.

Re: Generative AI coding tools and agents do not work for me

#432

Post by icedchai »

Eisenstein wrote: post #387
perrygeo wrote: post #379

Earlier quoted context omitted.

> Getting answers to those question is learning, regardless of where the answer comes from. Sort of. The process of working through the question is what drives learning. If you just receive the answer with zero effort, you are explicitly bypassing the brain's learning mechanism. There's huge difference between your workflow and fully Agentic AIs though. Asking an AI for the answer in the way you describe isn't exactl…

The logical conclusion of this is 'the AI just solves the problem by coding without telling you about it'. If we think about 'what happens when everyone vibe-codes to solve their problems' then we get to 'the AI solves the problem for you, and you don't even see the code'. Vibe-coding is just a stop on the road to a more useful AI and we shouldn't think of it as programming.

It "tells you about it" with code. You can still learn from the code AI has produced. It may be suboptimal or messy... but so is code produced by many of our fellow humans.

Re: Generative AI coding tools and agents do not work for me

#433

Post by iLemming »

wiseowise wrote: post #161
handfuloflight wrote: post #6

Will we be having these conversations for the next decade?

It’s the new “I use Vim/Emacs/Ed over IDE”.

Jokes on you. LLMs integrate into Emacs so seamlessly, you probably have no idea. I can ask LLMs to help me at any point, whether I'm writing some notes, sending a Slack message to a colleague, editing a comment in a codebase or a git commit message, or even when running shell commands. You can easily manipulate the context applied to the conversation, see the payload, repeat with variability, swap models anytime, call external tools, replace things in place, examine the diff of the changes, search through your prior conversations, etc.

Re: Generative AI coding tools and agents do not work for me

#434

Post by computably »

wiseowise wrote: post #427
didibus wrote: post #414

Earlier quoted context omitted.

I think we are all working without data here, it's all conjecture. I went with OP's hypothesis that you are not faster, you throw things at the wall, wait, and see if it sticks, or re-throw it until it does. This reduces your cognitive load, but might not actually make you more productive. I'm assuming here that "you are not more productive" already accounted for what you are saying. Like in a 8h day, without AI, you…

> What you hypothesize could also be true, it the mental load is reduced, can you sustain a higher productivity for longer? We don't know, maybe. It's not maybe, it's confirmed fact. Otherwise there wouldn't be burnout epidemic.

Except the causes of burnout have almost nothing to do with the type of cognitive load associated with coding, debugging, etc.

https://www.mayoclinic.org/healthy-lifestyle/adult-health/in...

Of the six general causes listed, four are institutional or social, having to do more with the workplace or coworkers: lack of control, lack of clarity, interpersonal conflicts, lack of support. IME, in tech, these are far more common causes and more deeply tied to the root of the issue than specifics of work.

The remaining two are productivity-related issues: too much/little to do, problems with WLB.

I would note these are tied into lack of control/clarity/support, and conflict. In a healthy work environment, expectations should be clear and at least somewhat flexible depending on employee feedback, and adequate support should be provided by the employer.

That aside, it's unclear, and I would argue unlikely, that AI-related productivity gains will help with workload issues. If you do disproportionately more work in an overworked team/org, you will simply be given more work. If many people see gains in productivity, then either the bar for productivity goes up, or there's layoffs. Even if you manage to squeak by / quiet quit with much reduced cognitive load for coding, and that's most of your job, unless you are fully remote the most likely change is your butt-in-seat time will go from "mentally taxing coding" to "mentally toxic doomscrolling."

Re: Generative AI coding tools and agents do not work for me

#435

Post by swader999 »

bevr1337 wrote: post #121
throwawaysleep wrote: post #81

Earlier quoted context omitted.

If you lack "that kind of scratch", you are at the learning stage for software development, not the keeping up stage. Either that or horribly underpaid.

I recently had a coworker tell me he liked his last workplace because "we all spoke the same language." It was incredible how much he revealed about himself with what he thought was a simple fact about engineer culture. Your comment reminds me of that exchange. - Employers, not employees, should provide workplace equipment or compensation for equipment. Don't buy bits for the shop, nails for the foreman, or Cursor fo…

I agree with all this but the simple fact is that if you don't keep up you'll be out of a job faster than the rest of us. My strategy for being replaced by AI is to replace the company that replaces me. Software is getting trivial to implement, especially if you know how to specify it.

Re: Generative AI coding tools and agents do not work for me

#436

Post by swader999 »

hatefulmoron wrote: post #57
adriand wrote: post #28

Earlier quoted context omitted.

Do you have to review the code? I’ll be honest that, like the OP theorizes, I often just spot review it. But I also get it to write specs (often very good, in terms of the ones I’ve dug into), and I always carefully review and test the results. Because there is also plenty of non-AI code in my projects I didn’t review at all, namely, the myriad open source libraries I’ve installed.

It depends on what you're doing. If it's a simple task, or you're making something that won't grow into something larger, eyeballing the code and testing it is usually perfect. These types of tasks feel great with Claude Code. If you're trying to build something larger, it's not good enough. Even with careful planning and spec building, Claude Code will still paint you into a corner when it comes to architecture. In…

Vertical slice architecture keeps you out of the corner.

Re: Generative AI coding tools and agents do not work for me

#437

Post by skydhash »

dvt wrote: post #431
skydhash wrote: post #309

Earlier quoted context omitted.

For this case, i will probably lift off the example from the library docs. Or spend 5 minutes writing a bare implementation as it would be all I need at the time. That’s an issue I have with generated code. More often, I start with a basic design that evolves based on the project needs. It’s an iterative process that can span the whole timeline. But with generated code, it’s a whole solution that fits the current nee…

> For this case, i will probably lift off the example from the library docs. Or spend 5 minutes writing a bare implementation as it would be all I need at the time. Both of these would take longer than 5 minutes. There's also no "lifting" as this case involves both Provider and Context, so you'd have to combine React doc examples. The only alternative would be knowing it by heart, which you might, but I don't (nor do…

What about understanding. I've not touched React for some times, but I'm familiar enough with the library to locate every piece of exact information I need. And this for a lot of framework, languages, and libraries I've used over the year. There's documentation browsers like dash.app, zeal, devdocs, and various cheat sheets that help.

Re: Generative AI coding tools and agents do not work for me

#438

Post by genewitch »

danielbln wrote: post #319
rsanheim wrote: post #249

Earlier quoted context omitted.

It's just Pattern matching. Most APIs, and hell, most code is not unique or special. Its all been done a thousands of times before. Thats why an LLM can be helpful on some tool you've written just for yourself and never released anywhere. As to 'knows the answer', I'm don't even know what that means with these tools. All I know is if it is helpful or not.

Also, most problems are decomposable into simpler, certainly not novel parts. That intractable unicorn problem I hear so much about is probably composed of very pedestrian sub-problems.

to what does "unicorn problem" refer to? a specific thing or a general idea?

Re: Generative AI coding tools and agents do not work for me

#439

Post by genewitch »

layer8 wrote: post #262
sagarpatil wrote: post #87

Earlier quoted context omitted.

I always use Claude Code to debug issues, there’s no point in trying to do this yourself when AI can fix it in minutes (easy to verify if you write tests first) o3 with new search can do things in 5 mins that will take me at least 30 mins if I’m very efficient. Say what you want but the time savings is real.

Tests can never verify the correctness of code, they only spot-check for incorrectness.

"hey claude, does this function terminate?"

Re: Generative AI coding tools and agents do not work for me

#440

Post by genewitch »

materielle wrote: post #154
didibus wrote: post #132

You could argue that AI-generated code is a black box, but let's adjust our perspective here. When was the last time you thoroughly reviewed the source code of a library you imported? We already work with black boxes daily as we evaluate libraries by their interfaces and behaviors, not by reading every line. The distinction isn't whether code comes from AI or humans, but how we integrate and take responsibility for i…

I’m not sure that the library comparison really works. Libraries are maintained by other humans, who stake their reputation on the quality of the library. If a library gets a reputation of having a lax maintainer, the community will react. Essentially, a chain of responsibility, where each link in the chain has an incentive to behave well else they be replaced. Who is accountable for the code that AI writes?

> Who is accountable for the code that AI writes?

i say we make it the original publishers of the data ingested by the AI during training. Just for the court battles.

Post reply on HN