> Whenever I find a use for AI, it so drastically cuts the time to accomplish tasks that I’m left wanting to do more with my (finally copious) free time!
I get what the author is trying to say but this does not match my experience nor does it match the employee productivity findings from AI adoption in enterprises.
Yeah, the hidden implication here is that they do not meaningfully review or design what the AI creates. Which is not actually an invalid AI use case, but it is highly misleading to leave it unstated.
(And of course there is the uncharitable alternative - that they are simply incompetent. But we don't actually need to go there.)
In my experience, using chats merely as stack overflow replacement, e.g. for providing obscure API usage examples, is a huge time saver — otherwise I would’ve had to lurk around game engine sources, examples, comments, sometimes forum posts, etc. AI code reviews definitely help finding issues, especially when I work on something for more than a week and inevitably forgetting things.
This is my experience. It's a huge time saver for learning something that you would have had to previously Google or wade through multiple pages of documentation for, as long as you develop an intuition for when to fact-check it. It's a slightly less huge time saver for one-shotting simple functions/scripts ("write a function to download everything in a given time range from this paginated API and save it to files"), assuming you actually review what it outputs. But neither of those things are where most time is spent in software.
> It's a huge time saver for learning something that you would have had to previously Google or wade through multiple pages of documentation for, as long as you develop an intuition for when to fact-check it.
When you're learning something, wading through documentation is a lot more valuable than zeroing in on the one thing faster. It's an investment in the future.
You don't need to do this if you setup TDD. Over time, you develop confidence in the tests, and those ensure your code is solid. Linters/formatters ensure coding standards.
It's a shift, but once you make it, is when agentic development clicks.
This is true but it's completely negated when someone wastes my time asking them to review their AI slop wall of code, or when it actually goes live and blows up production.
I was going to say that this surely depends on the type of work being done with AI but then a small dive into your HN profile shows you're working on SWE agents at Google?
I am struggling to reconcile your viewpoint with this but may have found the wrong profile!
It's possible to be frank about the misuse of AI in a lot of corporate settings today and also view it as a tool (or a step in tool development) that has the potential to be quite valuable.
Also, there are plenty of non-development related uses of AI that are genuinely valuable - document extraction and data repair are strong domains for application.
Absolutely agreed - though the usual claim for coding agents in particular is that they are already very capable and fast, replacing and speeding up software development hugely (anecdotally reasonably true for me and my colleagues, though again depends on domain and complexity).
Author here. I don't use it for everything and I don't expect it to do everything. I've used it for image generation, code generation, and also iterated on prototypes / simple visual simulations in languages and frameworks that I (mostly) understand.
And, of course, like everyone else, I've used it instead spending a lot more time searching and sifting through things online. I used to think replacing the search bar with an AI prompt was gimmicky, but then the replies improved.
Those use-cases alone have been massive time savers for me.
> Swimming against the rip will leave you exhausted and helpless. If you are though, then I hope you have a large life preserver.
You use AI for search like everyone else because that's what it always was and all it was ever good for. What if I told you that your dramatic statement is totally misreading the vast majority of office workers and their work? If there are waves, it doesn't resemble a riptide, but stop-and-go traffic.
Their work isn't hard for them to begin with. They're bored and stuck in endless meetings. The bottleneck is the rest of the organization. Just like a traffic jam, they cannot leave. Not even working from home will save them from the random calls, messages, or emails keeping them glued to the desk.
For every nontrivial goal, all decisions up to that point will await approval from one person. The baton is then handed off to someone else and so it goes.
As tempting as it is for some people to want to, one cannot jam the AI middleman pirate grift machine into this process without violating centuries of accountability laws. Anyone attempting to build an AI-first business will hit these same roadblocks and become the thing they attempted to be "better" than.
Any attempt by AI companies to slip out of this situation with reform politics would be highly entertaining to watch though. I think the sharks smell blood.
I think you're right about companies and bottlenecks. I don't think it disagrees with what I'm saying.
The rip is in the demand and adoption of the technology and its subsequent effects on everything else. You can choose to learn how to navigate it, go faster with it, or sit on the sidelines, but trying to stop it won't make the demand, the technology or its adoption go away.
This thing is here to stay and , judging by all the meetups in my area, it's only getting bigger. Who knows, maybe companies will be forced to dramatically change to reduce bottlenecks. Maybe there will be a lot more smaller companies. I can't say for sure.
The interest of businesses and their employees do not control how they are required to conduct business. The tail does not wag the dog.
This is an uphill legal battle beyond personal opinion or preference, and I'm not seeing the demand. I'm not saying that you think vibecoding is what people should be doing with AI, but ignoring the details is the only way to make this go faster. External to the business groupthink, that is just fraud.
All this time before LLMs, any business could have just cranked out slop any old way. What kept them from doing that?
What I think AI has done is shine a spotlight on longstanding problems, and that is good, but we're a lot further away from solving any of them than many realize. I'm not even sure there is anything to solve if we're just talking about efficiency.
> The rip is in the demand and adoption of the technology and its subsequent effects on everything else. You can choose to learn how to navigate it, go faster with it, or sit on the sidelines, but trying to stop it won't make the demand, the technology or its adoption go away.
What is there to adapt to? I don't think AI has changed work. It improved accessibility to information. It didn't exceed the capability of web search or books. You still have to type some stuff and then read what it "generates". That's... that's the same thing we were already doing.
> The barrier to entry for a creative person to benefit from AI is nothing.
The barrier is low, but the benefit is negligible or even negative. You're outsourcing your creative process to some mega-corporation. What you get back from it might be interesting to you, but it won't be to a lot of other people except maybe as a novelty, because it's from the same megacorp that sells its services to everyone else. I know some people seem to genuinely enjoy the outputs of the AI models now, but the novelty hasn't worn off yet.
Also, congratulations, you get to live with the stigma of being an AI slop pusher. And if you try to pass the work off as your own, good luck hiding the truth for the rest of your life, because you'll never regain trust if people find out.
- I think we’ve all learned with social media that sometimes tech companies create products with good intentions that are later weaponized against their users.
- The foundation of agentic LLMs is fully precipitated on creative and knowledge theft. They stole our bread and now want to sell it back to us, one token at a time.
- Banning AI has an intention to slow down its adoption while society weighs in on its projected consequences. When did we as a society collectively agree tech entrepreneurs get to release anything they want without any oversight?
This “just accept it” attitude is democratically lazy and so very ignorant of history of what rich people in power tend to do to the working class when they don’t need their labor anymore. There’s blood in the machine, friends.
Some people want to fully ban AI and put the genie back in the bottle[1] but I'd much rather it simply be demystified and genuinely costed.
I think most people would prefer to find a useful application of a tool compared to being compelled by corporate overlords to clock in at a certain level of token spend. It isn't magic or a cure-all - it's something that's very useful sometimes and insanely expensive (in dollar cost and wasted effort of real humans) other times. Right now we seem to be in a weird honeymoon phase where it gets used for everything disregarding how much utility or cost it will have in that setting.
1. It's impossible because for decades we've had plenty of neural network applications and other ML uses that are quite useful.
Banning smoking from indoor spaces did in fact make it go away [from indoor spaces].
Actually, the analogy is compelling - an addictive habit that is pleasurable to the user, profitable for the supplier, but others around them must suffer the effluent with consequent deleterious effects on [mental] health. And similarly, the worst [mental] health effect is of course on the user themselves.
As with smoking, probably we can't/shouldn't stop individuals from indulging in this habit, but surely we can protect the rest of society.
AI is very useful in the right hands. But this is a great example of a destructive attitude from which I earnestly pray most engineers escape. This is, in my opinion, the kind of person who is destroying the world I love so much.
Frankly, as much as I would prefer an in-between solution, it might not be feasible. If I could press a button that made AI disappear forever, I would, despite the sacrifice. The downsides as compared to the upsides, in practice, are just so overwhelming.
> Whenever I find a use for AI, it so drastically cuts the time to accomplish tasks that I’m left wanting to do more with my (finally copious) free time!
I get what the author is trying to say but this does not match my experience nor does it match the employee productivity findings from AI adoption in enterprises.
Yeah, the hidden implication here is that they do not meaningfully review or design what the AI creates. Which is not actually an invalid AI use case, but it is highly misleading to leave it unstated.
(And of course there is the uncharitable alternative - that they are simply incompetent. But we don't actually need to go there.)
In my experience, using chats merely as stack overflow replacement, e.g. for providing obscure API usage examples, is a huge time saver — otherwise I would’ve had to lurk around game engine sources, examples, comments, sometimes forum posts, etc. AI code reviews definitely help finding issues, especially when I work on something for more than a week and inevitably forgetting things.
This is my experience. It's a huge time saver for learning something that you would have had to previously Google or wade through multiple pages of documentation for, as long as you develop an intuition for when to fact-check it. It's a slightly less huge time saver for one-shotting simple functions/scripts ("write a function to download everything in a given time range from this paginated API and save it to files"), assuming you actually review what it outputs. But neither of those things are where most time is spent in software.
> It's a huge time saver for learning something that you would have had to previously Google or wade through multiple pages of documentation for, as long as you develop an intuition for when to fact-check it.
When you're learning something, wading through documentation is a lot more valuable than zeroing in on the one thing faster. It's an investment in the future.
> assuming you actually review what it outputs
You don't need to do this if you setup TDD. Over time, you develop confidence in the tests, and those ensure your code is solid. Linters/formatters ensure coding standards.
It's a shift, but once you make it, is when agentic development clicks.
Yeah, AI works great for everyone in vacuum, but works for no one in a group.
This is true but it's completely negated when someone wastes my time asking them to review their AI slop wall of code, or when it actually goes live and blows up production.
I was going to say that this surely depends on the type of work being done with AI but then a small dive into your HN profile shows you're working on SWE agents at Google?
I am struggling to reconcile your viewpoint with this but may have found the wrong profile!
It's possible to be frank about the misuse of AI in a lot of corporate settings today and also view it as a tool (or a step in tool development) that has the potential to be quite valuable.
Also, there are plenty of non-development related uses of AI that are genuinely valuable - document extraction and data repair are strong domains for application.
Absolutely agreed - though the usual claim for coding agents in particular is that they are already very capable and fast, replacing and speeding up software development hugely (anecdotally reasonably true for me and my colleagues, though again depends on domain and complexity).
Author here. I don't use it for everything and I don't expect it to do everything. I've used it for image generation, code generation, and also iterated on prototypes / simple visual simulations in languages and frameworks that I (mostly) understand.
And, of course, like everyone else, I've used it instead spending a lot more time searching and sifting through things online. I used to think replacing the search bar with an AI prompt was gimmicky, but then the replies improved.
Those use-cases alone have been massive time savers for me.
> Swimming against the rip will leave you exhausted and helpless. If you are though, then I hope you have a large life preserver.
You use AI for search like everyone else because that's what it always was and all it was ever good for. What if I told you that your dramatic statement is totally misreading the vast majority of office workers and their work? If there are waves, it doesn't resemble a riptide, but stop-and-go traffic.
Their work isn't hard for them to begin with. They're bored and stuck in endless meetings. The bottleneck is the rest of the organization. Just like a traffic jam, they cannot leave. Not even working from home will save them from the random calls, messages, or emails keeping them glued to the desk.
For every nontrivial goal, all decisions up to that point will await approval from one person. The baton is then handed off to someone else and so it goes.
As tempting as it is for some people to want to, one cannot jam the AI middleman pirate grift machine into this process without violating centuries of accountability laws. Anyone attempting to build an AI-first business will hit these same roadblocks and become the thing they attempted to be "better" than.
Any attempt by AI companies to slip out of this situation with reform politics would be highly entertaining to watch though. I think the sharks smell blood.
I think you're right about companies and bottlenecks. I don't think it disagrees with what I'm saying.
The rip is in the demand and adoption of the technology and its subsequent effects on everything else. You can choose to learn how to navigate it, go faster with it, or sit on the sidelines, but trying to stop it won't make the demand, the technology or its adoption go away.
This thing is here to stay and , judging by all the meetups in my area, it's only getting bigger. Who knows, maybe companies will be forced to dramatically change to reduce bottlenecks. Maybe there will be a lot more smaller companies. I can't say for sure.
The interest of businesses and their employees do not control how they are required to conduct business. The tail does not wag the dog.
This is an uphill legal battle beyond personal opinion or preference, and I'm not seeing the demand. I'm not saying that you think vibecoding is what people should be doing with AI, but ignoring the details is the only way to make this go faster. External to the business groupthink, that is just fraud.
All this time before LLMs, any business could have just cranked out slop any old way. What kept them from doing that?
What I think AI has done is shine a spotlight on longstanding problems, and that is good, but we're a lot further away from solving any of them than many realize. I'm not even sure there is anything to solve if we're just talking about efficiency.
> The rip is in the demand and adoption of the technology and its subsequent effects on everything else. You can choose to learn how to navigate it, go faster with it, or sit on the sidelines, but trying to stop it won't make the demand, the technology or its adoption go away.
What is there to adapt to? I don't think AI has changed work. It improved accessibility to information. It didn't exceed the capability of web search or books. You still have to type some stuff and then read what it "generates". That's... that's the same thing we were already doing.
the goal of proselytism is always the same -- to help the proselytizer convince themselves they are not the mark
> productivity findings
are you referring to the early 2025 METR study?
It will just make it so only the privileged can benefit - and that list will not include you.
> The barrier to entry for a creative person to benefit from AI is nothing.
The barrier is low, but the benefit is negligible or even negative. You're outsourcing your creative process to some mega-corporation. What you get back from it might be interesting to you, but it won't be to a lot of other people except maybe as a novelty, because it's from the same megacorp that sells its services to everyone else. I know some people seem to genuinely enjoy the outputs of the AI models now, but the novelty hasn't worn off yet.
Also, congratulations, you get to live with the stigma of being an AI slop pusher. And if you try to pass the work off as your own, good luck hiding the truth for the rest of your life, because you'll never regain trust if people find out.
Bad take, if you ask me, on several points:
- I think we’ve all learned with social media that sometimes tech companies create products with good intentions that are later weaponized against their users.
- The foundation of agentic LLMs is fully precipitated on creative and knowledge theft. They stole our bread and now want to sell it back to us, one token at a time.
- Banning AI has an intention to slow down its adoption while society weighs in on its projected consequences. When did we as a society collectively agree tech entrepreneurs get to release anything they want without any oversight?
This “just accept it” attitude is democratically lazy and so very ignorant of history of what rich people in power tend to do to the working class when they don’t need their labor anymore. There’s blood in the machine, friends.
Some people want to fully ban AI and put the genie back in the bottle[1] but I'd much rather it simply be demystified and genuinely costed.
I think most people would prefer to find a useful application of a tool compared to being compelled by corporate overlords to clock in at a certain level of token spend. It isn't magic or a cure-all - it's something that's very useful sometimes and insanely expensive (in dollar cost and wasted effort of real humans) other times. Right now we seem to be in a weird honeymoon phase where it gets used for everything disregarding how much utility or cost it will have in that setting.
1. It's impossible because for decades we've had plenty of neural network applications and other ML uses that are quite useful.
Of course conformity is the easiest choice. But don’t lie to yourself that it’s the best choice.
Banning smoking from indoor spaces did in fact make it go away [from indoor spaces].
Actually, the analogy is compelling - an addictive habit that is pleasurable to the user, profitable for the supplier, but others around them must suffer the effluent with consequent deleterious effects on [mental] health. And similarly, the worst [mental] health effect is of course on the user themselves.
As with smoking, probably we can't/shouldn't stop individuals from indulging in this habit, but surely we can protect the rest of society.
[dead]
AI is very useful in the right hands. But this is a great example of a destructive attitude from which I earnestly pray most engineers escape. This is, in my opinion, the kind of person who is destroying the world I love so much.
Frankly, as much as I would prefer an in-between solution, it might not be feasible. If I could press a button that made AI disappear forever, I would, despite the sacrifice. The downsides as compared to the upsides, in practice, are just so overwhelming.
[dead]
I’m impressed with how vapid a post can be and still find success on Hacker News.