You’re right I have no idea, I use it for scripts, it works well, you guys are all linux nerds talking about philosophy that I don’t care for
I need code to make a script, it works, end of story, I don’t care about your philosophy on less code more code better code worse code, and I’m pretty sure most businesses don’t care either
I’m a C++ programmer most of the time, but I also write scripts. Claude and even Gemini write better bash scripts than I do - better error handling, better commenting, better input sanitizing, better use of parameters - because, all that bash syntax is an annoying pile of illogical junk that I have to look up every time I go to use it, and what are AI agents really fast at doing? Looking stuff up.
sure let’s just burn down half a hectare of the amazon for a seven line script.
engineering philosophy is where the rubber hits the road. building software with the right philosophy can mean the difference between needing a datacenter or a raspberry pi for the same job. it directly translates to money saved in recurring costs.
The seven line script isn’t burning down the rainforest. Programmers are a 1% slice of AI token usage. A seven line script takes less datacenter power to generate than your monitor does to show you this message thread while you read it. What’s burning down the rainforest is billions of “ordinary” people using AI to make Rule 34 animations, graphic layout flyers for their next coffee date, research papers on obscure topics that will never be read by anyone, schemes to trade bitcoin, etc. etc. etc.
I glanced across one once, I doubt any of the demographics are entirely reliable, but even with the uncertainty it appears that programmers are but a small minority even among Claude users.
to be fair I use https://www.ecosia.org/ which is climate positive so anything else I do on the web is carbon negative, however I also have solar and a battery and export power at night so I am technically climate positive (I reduce more emissions than I make)
also I mainly use lechat tbh, it’s french and aims to be climate friendly:
The project will support large-scale AI model training and inference workloads, leveraging Sweden’s renewable energy capacity and EcoDataCenter’s established footprint in sustainable data center operations. The facility is expected to be powered primarily by renewable energy sources, aligning with both companies’ commitments to reducing the carbon intensity of AI compute.
If the grid is more green, the data centres are more green, and in the most tech heavy place on earth:
The growing portfolio of grid-scale batteries in California, the world’s fourth biggest economy, hit a stunning new peak earlier this week, reaching a share of 44 pct of evening demand at once stage in the early evening.
It’s not my fault Americans are innovative enough to come up with a magic computer code box but not smart enough to install a bunch of solar panels and batteries :\
Both. So, if you - or other meatbags - write the code, AI agents can review it faster / more thoroughly (therefore: better) than human reviewers, finding more problems and proposing fixes.
If you accept every fix your reviewer proposes, human or AI, you’re an idiot. But if you ignore every bug a human or AI reviewer raises as a possibility, you’re an even bigger idiot. If your code has no bugs, you’re a liar.
You can always tell who the vibe/shit/non coders are by the way they pontificate AI as some revolution that completely changes software, instead of yet another tool that has significant tradeoffs…
I’ll take that as a sign you’ve never written a single line of code in your life, since if you actually did so, you’d soon realise it’s like having a fast typing, cocaine fuelled intern as a sidekick, who only saw a couple stack overflow code snippets in your language of choice and ran with it.
I find a lot of the AI code hate coming from people who apparently have never managed software production by less-than-perfect programmers. If you handle AI code like it’s coming from a team of consultants on a revolving-door staffing arrangement, it’s about like that. Ideal? No. If you hire 9 women can you get them to have your baby within one month? No. But if you are handed 10 programmers who know almost nothing about your specific project, you generally can get them to produce useable code at about the rate of 3 programmers who are familiar with the project - iff you’re disciplined about everything they do.
Then maybe there’s something here to learn from for you. Or don’t and assume that your vibe coding skills are the same as career engineers who’ve dedicated their lives to writing quality systems and producing beautiful code?
If it speeds up your workflow by that much (and I’ll admit it does make some tasks trivial for me now) that’s way more telling about your current skillset than the general state of the art.
That’s kind of a point for it? If the code isn’t the product, then it doesn’t matter who made it if the product is built properly. And sota models certainly know where to put the nails.
A better analogy would be a 3d printer. Much quicker than carving stuff out of wood. Might fail a print but it’s still faster. It’s not exactly food safe so avoid using it for something important. Obviously, it works well for small things but trying to build a house out of 3d printed bricks is going to yield failure.
Not the best analogy though but nails and hammers and monkeys is just missing the point completely imo.
Not sure what you mean by this tbh. The client list in relation to the product? Do you mean the source code leak? That’s a bit what I mean by it not being safe for everything and trying to build a house out of 3d printed bricks. I’m not even certain the code leak was caused by AI generated code anyways.
I personally only build small internal apps for the rest of the team. We don’t ship software.
Granted, every tool has its use and most can be used for the wrong thing if you try hard enough. It’s also easier to use it for the wrong thing with AI, has to be said.
think of it like this. you build small tools for internal use, so code quality, maintainability and documentation are not your highest priorities, right? most important is to ship the features needed by your colleagues right now. i’ve been in that same boat, building internal testing tools for a big multinational.
say your tool contains shortcuts for interacting with some internal database. you don’t need auth because it’s all on the internal net, so it’s just a collection of shortcuts. you don’t really care about maintainability, because it’s all just temporary, so you throw every new request together in the fastest way you can, probably trying out new techniques to keep yourself entertained. you don’t really care about testability, because you’re the only one testing so you can check that everything works before doing a release and if something slips through one of your colleagues will walk down the hall and tell you.
now imagine it gets enough attention that your boss says “we want our customers to use this”. suddenly your priorities are upended: you absolutely need auth, you definitely need testability and you absolutely definitely need the tool to not mess things up. best possible world, you can reimplement the tool in a more manageable way, making sure every interface is consistent, documentation is up to snuff for users, and error handling is centralised.
the claude cli leak is the opposite of that. it’s the worst code quality i’ve ever seen. it’s full of errors, repeated code, and overzealous exception handling. it is absolutely unmaintainable. the functionality for figuring out what type a file is and reading it into a proper object is 38 000 lines of typescript, excluding the class definitions. the entire thing is half a million lines. the code for uploading a pdf calls the code to upload jpegs, because if a file isn’t identified it’s automatically a jpeg. and jpegs can go through the same compression routine up to 22 times before it tries to upload them because the handler just calls itself repeatedly.
and this is the code they thought was robust enough to withstand the internet. imagine what their internal tooling looks like.
because it’s all just temporary, so you throw every new request together in the fastest way you can
I have been finding many use cases for “throwaway” AI coded tools like parser based calculators… the old: “if it takes you a day to write it and it speeds up a week long task by even 20% then it’s a win” thing has had the cost of “writing it” come down by a big factor for simple things - a parser tool thingy that might have taken 4 hours can now be certified good enough for internal use in 40 minutes, and if it’s saving 3 hours of hand work that has turned a “meh, maybe we’ll do this again sometime - I’d rather write the tool than drudge through 3 hours of brain-dead error prone work” into an overall big win.
i will never understand that attitude. four hours of exploration, learning and puzzle-solving sounds like the best part of the job. an isolated, well-specified problem that can be completed in a day is like the most fun you can have programming. why swap that for an hour of code review?
I guess it’s not puzzle solving for me anymore. The puzzle is solved in my head, the four hours of exploration are mostly occupied with looking up annoying details about how to do what’s in my head within the syntax and limitations of the latest re-invention of whatever lanugage has been selected / is available in the environment. Both routes have to be tested, but glancing over provided source code and nodding “yeah, that looks right” then testing the successful implementation is a hell of a lot more satisfying for me than deciphering complier errors, untangling version incompatibilities, looking up function signatures, etc.
Speed and quality are basically uncorrelated, except in the big picture where slower -> better -> fastest to a real solution for the whole system.
The reason slower leads to better is due to iterations of examination and consideration of more possible use cases. It’s not about wall clock time, it’s about understanding the problem. AI agents let you shoot yourself in the foot 10x faster, but they also help you work through the edge cases faster too - not 10x faster, because the agent can’t know the edge cases as well as a domain expert can, but still faster.
It can be rather helpful for some things but I’ve yet to see any actual data supporting it being “substantially faster”. To make something that works and is of equal quality, that is.
I mean if you’re just making the same slightly different website for different customers or similarly boilerplate stuff, I’m sure it’s substantial. But that’s not really most of dev.
Yeah i duno, it’s helped me make a massive script to automate a hugely time consuming part of my job saving literally 10’s of thousands of hours of work, I’m using it to also setup a boilerplate website that links to our backend to monitor services, it has also helped me setup home automation which makes me money (export power to grid during peak time), and I use it daily for any number of things, it’s great
Claude Code hit $1 billion in annualized run-rate revenue in the first six months after its release and has since grown to $2.5 billion, the company said. Once used primarily by AI-forward startups, Claude Code has gained traction with engineering teams at Fortune 500 companies and even among hobbyists lacking technical skills who are interested in building their own apps. It’s been used for everything from growing a tomato plant to helping plan the route of a NASA Mars rover. On social media, users describe themselves as “Claude-pilled,” or Claude-obsessed.
3D printing was awesome. Once or twice a year, it could make something for me that was genuinely useful. It also occupied a big chunk of space in the house and mostly created mostly useless plastic junk creating more clutter - we started 3D printing in 2018 and gave the printer away to someone with a bigger house in 2020. Once since then I have asked him to print something for me. It’s that useful.
AI agents are making software. I’ve got Terabyte flash drives smaller than a key fob - the clutter won’t be a problem in our house anytime soon. So far, Claude has helped me cleanup / fix several hobby software projects in amazingly short time to an amazingly high level of polish.
I just used Claude yesterday to add some functionality to an existing python script that interacts with AWS. It created an unnecessary loop where 2 of 3 iterations were effectively no-ops. So while it did ultimately provide what I needed, I still had to refactor what it generated in order to remove the useless loop.
Oh, about the un-necessary loops thing. I was once hired by a PhD to build a software development department to convert his “academic matlab code” into an actual commercial application. One of my programmers was another PhD, and he successfully translated the matlab into C++ and it did run faster, and he parallelized it successfully (the latest hotness in 2005) but was still the slowest part of our whole operation. So, we did an old-fashioned code review and found an un-necessary loop, unwound that, got the exact same answers 100x faster. Apparently the PhDs were loop-slop blind, one had missed it for years and the other had been working on the translation/optimization/parallelization for months without seeing it.
Remember, it was trained on Stack Overflow… a self-selected collection of “beginner” questions with a lot of “this is good enough to get you unstuck” answers. If you want better, you have to specify and test to ensure you get better.
The outages have been so brief, and the speed margin is so great now, that waiting for the outages is a trivial delay as compared to “doing it the old fashioned way.”
Claude is used mainly for writing code, it is substantially faster at writing code than a human
Question is do we need more code, or better code?
The key to successful use of AI is: use it to write better code, not more code. This means iterations - just like when humans write code.
Who are you asking that question to? I don’t care about your code, make it as shitty as you want
Less code is better code. And you’re response indicates you do not understand that point.
You’re right I have no idea, I use it for scripts, it works well, you guys are all linux nerds talking about philosophy that I don’t care for
I need code to make a script, it works, end of story, I don’t care about your philosophy on less code more code better code worse code, and I’m pretty sure most businesses don’t care either
I’m a C++ programmer most of the time, but I also write scripts. Claude and even Gemini write better bash scripts than I do - better error handling, better commenting, better input sanitizing, better use of parameters - because, all that bash syntax is an annoying pile of illogical junk that I have to look up every time I go to use it, and what are AI agents really fast at doing? Looking stuff up.
🤣 so true!
I picked up python fairly quickly but bash my gosh
sure let’s just burn down half a hectare of the amazon for a seven line script.
engineering philosophy is where the rubber hits the road. building software with the right philosophy can mean the difference between needing a datacenter or a raspberry pi for the same job. it directly translates to money saved in recurring costs.
The seven line script isn’t burning down the rainforest. Programmers are a 1% slice of AI token usage. A seven line script takes less datacenter power to generate than your monitor does to show you this message thread while you read it. What’s burning down the rainforest is billions of “ordinary” people using AI to make Rule 34 animations, graphic layout flyers for their next coffee date, research papers on obscure topics that will never be read by anyone, schemes to trade bitcoin, etc. etc. etc.
i’ve not actually looked at the demographics. is that available somewhere?
I glanced across one once, I doubt any of the demographics are entirely reliable, but even with the uncertainty it appears that programmers are but a small minority even among Claude users.
that would probably be a couple hundred tokens so the same energy use as a google search
a google ai search, sure. a normal web search uses several orders of magnitude less energy.
to be fair I use https://www.ecosia.org/ which is climate positive so anything else I do on the web is carbon negative, however I also have solar and a battery and export power at night so I am technically climate positive (I reduce more emissions than I make)
also I mainly use lechat tbh, it’s french and aims to be climate friendly:
https://dcpulse.com/news/mistral-ai-ecodc-ai-facility-sweden
If the grid is more green, the data centres are more green, and in the most tech heavy place on earth:
https://reneweconomy.com.au/grid-batteries-reach-stunning-new-peak-of-44-pct-of-evening-demand-in-worlds-fourth-biggest-economy/
They’re rapidly getting pretty green
It’s not my fault Americans are innovative enough to come up with a magic computer code box but not smart enough to install a bunch of solar panels and batteries :\
Stop talking as if you give any semblance of a shit about the art and craft of programming. You clearly don’t.
upvoted because you’re right, i don’t care?
edit: javascript sucks btw i hate it, it can suck MY art and craft
It’s substantially faster than most programmers at finding bugs too.
*creating
Both. So, if you - or other meatbags - write the code, AI agents can review it faster / more thoroughly (therefore: better) than human reviewers, finding more problems and proposing fixes.
If you accept every fix your reviewer proposes, human or AI, you’re an idiot. But if you ignore every bug a human or AI reviewer raises as a possibility, you’re an even bigger idiot. If your code has no bugs, you’re a liar.
*finding
https://www.bleepingcomputer.com/news/security/claude-ai-finds-vim-emacs-rce-bugs-that-trigger-on-file-open/
more fuel for the vim-emacs holy war
You can always tell who the vibe/shit/non coders are by the way they pontificate AI as some revolution that completely changes software, instead of yet another tool that has significant tradeoffs…
“But mongo DB is web scale!?!” — dunning-krugers
If it writes a bunch of code, but it’s mostly slop and dogshit and introduces bugs at an alarming rate, does speed matter?
Hello! You are the 1908321094318709312th person to say this line today, here is your award 🍰
I’ll take that as a sign you’ve never written a single line of code in your life, since if you actually did so, you’d soon realise it’s like having a fast typing, cocaine fuelled intern as a sidekick, who only saw a couple stack overflow code snippets in your language of choice and ran with it.
I find a lot of the AI code hate coming from people who apparently have never managed software production by less-than-perfect programmers. If you handle AI code like it’s coming from a team of consultants on a revolving-door staffing arrangement, it’s about like that. Ideal? No. If you hire 9 women can you get them to have your baby within one month? No. But if you are handed 10 programmers who know almost nothing about your specific project, you generally can get them to produce useable code at about the rate of 3 programmers who are familiar with the project - iff you’re disciplined about everything they do.
ok? 👍
Then maybe there’s something here to learn from for you. Or don’t and assume that your vibe coding skills are the same as career engineers who’ve dedicated their lives to writing quality systems and producing beautiful code?
If it speeds up your workflow by that much (and I’ll admit it does make some tasks trivial for me now) that’s way more telling about your current skillset than the general state of the art.
i don’t get it what am i supposed to be learning?
I don’t? stop assuming im a professional coder, i’ve said multiple times what i use it for in this very thread
tell that to microsoft and salesforce and slack and jira and every other piece of junk software that was garbage before ai
if ai is fixing bugs in their shitty “beautifully coded” software so much the better
code is not the product. code is the tool we use to build the product. it’s not about swinging your hammer, it’s about knowing where to put the nails.
That’s kind of a point for it? If the code isn’t the product, then it doesn’t matter who made it if the product is built properly. And sota models certainly know where to put the nails.
A better analogy would be a 3d printer. Much quicker than carving stuff out of wood. Might fail a print but it’s still faster. It’s not exactly food safe so avoid using it for something important. Obviously, it works well for small things but trying to build a house out of 3d printed bricks is going to yield failure.
Not the best analogy though but nails and hammers and monkeys is just missing the point completely imo.
after the client leak i should think it was even more poignant. a bit like tearing down a wall and finding it full of mouldy razor blades
Not sure what you mean by this tbh. The client list in relation to the product? Do you mean the source code leak? That’s a bit what I mean by it not being safe for everything and trying to build a house out of 3d printed bricks. I’m not even certain the code leak was caused by AI generated code anyways.
I personally only build small internal apps for the rest of the team. We don’t ship software.
Granted, every tool has its use and most can be used for the wrong thing if you try hard enough. It’s also easier to use it for the wrong thing with AI, has to be said.
think of it like this. you build small tools for internal use, so code quality, maintainability and documentation are not your highest priorities, right? most important is to ship the features needed by your colleagues right now. i’ve been in that same boat, building internal testing tools for a big multinational.
say your tool contains shortcuts for interacting with some internal database. you don’t need auth because it’s all on the internal net, so it’s just a collection of shortcuts. you don’t really care about maintainability, because it’s all just temporary, so you throw every new request together in the fastest way you can, probably trying out new techniques to keep yourself entertained. you don’t really care about testability, because you’re the only one testing so you can check that everything works before doing a release and if something slips through one of your colleagues will walk down the hall and tell you.
now imagine it gets enough attention that your boss says “we want our customers to use this”. suddenly your priorities are upended: you absolutely need auth, you definitely need testability and you absolutely definitely need the tool to not mess things up. best possible world, you can reimplement the tool in a more manageable way, making sure every interface is consistent, documentation is up to snuff for users, and error handling is centralised.
the claude cli leak is the opposite of that. it’s the worst code quality i’ve ever seen. it’s full of errors, repeated code, and overzealous exception handling. it is absolutely unmaintainable. the functionality for figuring out what type a file is and reading it into a proper object is 38 000 lines of typescript, excluding the class definitions. the entire thing is half a million lines. the code for uploading a pdf calls the code to upload jpegs, because if a file isn’t identified it’s automatically a jpeg. and jpegs can go through the same compression routine up to 22 times before it tries to upload them because the handler just calls itself repeatedly.
and this is the code they thought was robust enough to withstand the internet. imagine what their internal tooling looks like.
I have been finding many use cases for “throwaway” AI coded tools like parser based calculators… the old: “if it takes you a day to write it and it speeds up a week long task by even 20% then it’s a win” thing has had the cost of “writing it” come down by a big factor for simple things - a parser tool thingy that might have taken 4 hours can now be certified good enough for internal use in 40 minutes, and if it’s saving 3 hours of hand work that has turned a “meh, maybe we’ll do this again sometime - I’d rather write the tool than drudge through 3 hours of brain-dead error prone work” into an overall big win.
i will never understand that attitude. four hours of exploration, learning and puzzle-solving sounds like the best part of the job. an isolated, well-specified problem that can be completed in a day is like the most fun you can have programming. why swap that for an hour of code review?
I guess it’s not puzzle solving for me anymore. The puzzle is solved in my head, the four hours of exploration are mostly occupied with looking up annoying details about how to do what’s in my head within the syntax and limitations of the latest re-invention of whatever lanugage has been selected / is available in the environment. Both routes have to be tested, but glancing over provided source code and nodding “yeah, that looks right” then testing the successful implementation is a hell of a lot more satisfying for me than deciphering complier errors, untangling version incompatibilities, looking up function signatures, etc.
Not just where to put the nails but also how hard to swing the hammer, when to stop swinging it, etc.
faster != better
In fact,
faster != good
Slow is smooth, smooth is fast
Speed and quality are basically uncorrelated, except in the big picture where slower -> better -> fastest to a real solution for the whole system.
The reason slower leads to better is due to iterations of examination and consideration of more possible use cases. It’s not about wall clock time, it’s about understanding the problem. AI agents let you shoot yourself in the foot 10x faster, but they also help you work through the edge cases faster too - not 10x faster, because the agent can’t know the edge cases as well as a domain expert can, but still faster.
works for me and apparently a shitload of other people because the thing is so popular it can’t stop crashing from people using the hell out of it
It can be rather helpful for some things but I’ve yet to see any actual data supporting it being “substantially faster”. To make something that works and is of equal quality, that is.
I mean if you’re just making the same slightly different website for different customers or similarly boilerplate stuff, I’m sure it’s substantial. But that’s not really most of dev.
If you’re reading a study performed 6, or even 3 months ago, you’re getting significantly stale data - it really is moving that fast.
Yeah i duno, it’s helped me make a massive script to automate a hugely time consuming part of my job saving literally 10’s of thousands of hours of work, I’m using it to also setup a boilerplate website that links to our backend to monitor services, it has also helped me setup home automation which makes me money (export power to grid during peak time), and I use it daily for any number of things, it’s great
Seems to work well for other people as well:
https://archive.md/4cBEV
But who knows, maybe it’s just 3d printing 2.0
3D printing was awesome. Once or twice a year, it could make something for me that was genuinely useful. It also occupied a big chunk of space in the house and mostly created mostly useless plastic junk creating more clutter - we started 3D printing in 2018 and gave the printer away to someone with a bigger house in 2020. Once since then I have asked him to print something for me. It’s that useful.
AI agents are making software. I’ve got Terabyte flash drives smaller than a key fob - the clutter won’t be a problem in our house anytime soon. So far, Claude has helped me cleanup / fix several hobby software projects in amazingly short time to an amazingly high level of polish.
I just used Claude yesterday to add some functionality to an existing python script that interacts with AWS. It created an unnecessary loop where 2 of 3 iterations were effectively no-ops. So while it did ultimately provide what I needed, I still had to refactor what it generated in order to remove the useless loop.
Oh, about the un-necessary loops thing. I was once hired by a PhD to build a software development department to convert his “academic matlab code” into an actual commercial application. One of my programmers was another PhD, and he successfully translated the matlab into C++ and it did run faster, and he parallelized it successfully (the latest hotness in 2005) but was still the slowest part of our whole operation. So, we did an old-fashioned code review and found an un-necessary loop, unwound that, got the exact same answers 100x faster. Apparently the PhDs were loop-slop blind, one had missed it for years and the other had been working on the translation/optimization/parallelization for months without seeing it.
Remember, it was trained on Stack Overflow… a self-selected collection of “beginner” questions with a lot of “this is good enough to get you unstuck” answers. If you want better, you have to specify and test to ensure you get better.
The outages have been so brief, and the speed margin is so great now, that waiting for the outages is a trivial delay as compared to “doing it the old fashioned way.”
Many monkeys can in fact swing hammers harder and faster than humans.
It doesn’t really make them better at it though
why are there so many lemmy users throwing weird analogies about monkeys and hammers at me
You’re supposed to be a smarter monkey?
I’m not sure
Fair