73 comments

  • danparsonson 2 hours ago

    Man selling shovels reports that every job requires a new shovel.

    Or in other words, of course he says that; it's not really a useful analysis is it?

    • rasz 2 hours ago

      Ray Kassar, CEO of Atari, says every Atari 2600 owner will buy at least 3 copies of E.T. the Extra-Terrestrial!

    • Fizzadar 2 hours ago

      100%. How is this even remotely considered newsworthy.

    • crinkly 2 hours ago

      Technically man selling shit and shovels. Or is it shovels that make shit.

  • 0xflarion 2 hours ago

    I remember 2018-2020, when everyone was saying that in 1-2 years, all cars would be autonomous vehicles and we won't need drivers anymore. Guess what.

    • flohofwoe an hour ago

      Same for UML in the 90s: "in a few years programmers will be obsolete because a handful system architects will draw UML diagrams which will be synthesized into code", in the end the opposite happened, and the dedicated system architect role pretty much disappeared.

      And UML even made a lot more sense to precisely describe a problem compared to human language prompts.

      • voxleone 39 minutes ago

        And using UML selectively and intentionally can build the habit of thinking before coding. It's a mental exercise, much like test-driven development or writing pseudocode, that strengthens your design intuition.

        • flohofwoe 33 minutes ago

          The question is if (too much) upfront planning is actually better.

          IME it's more useful to treat a source file (or new project) like a sketchbook, you start with an empty sheet of paper but already have a rough idea what you want, then you quickly sketch out the outlines, try out different ideas, explore different solutions (all without going into too much detail yet), you step through the code, explore some different paths to get an idea what "feels right", delete things, shuffle them around, rewrite them and slowly filling out the details and that way incrementally get towards the first working version.

          Which such an approach of incremental "micro-feedback-loops" you already eliminated a lot of dead ends that will appear anyway despite the best-laid plans - but identifying such dead ends early is much better than late.

          This incremental approach also forces you to keep the code small, tidy and malleable. The planning stage basically already happens in source code, and it's not one long planning stage, but many micro-planning-stages.

          ...this is also my problem with the current state of LLMs, they're pretty good at creating a first initial sketch from a very general problem statement (but only for things that have been done thousands of times by other people - but that's a different issue...) - but the more you'll need to go into the details and discover and fix problems in your initial 'mental design' (e.g. discovering what you actually wanted in the first place), the more the differences to traditional programming disappear, and at some tipping point it gets even more complex because you need to steer the LLM with a language (the human language) that simply isn't useful enough for detailed problem descriptions - that's why lawyers, math people and engineers all invented their own 'DSLs'.

          After a few iterations your prompts need to be just as detailed as writing source code in the first place - so what's the point of writing prompts in human language again?

          I feel like productivity could be improved much more by improving programming tools (yeah - boring old-school incremental maintenance work) instead of betting on some weird AI future which will just move the focus from writing in a programming-language to human-language prompts (which IMHO is a definitive step backwards because human language lacks precision - and adding that precision is how you end up with programming languages).

          Why are editing, compilation and debugging/testing still separate steps, why do debuggers still have those 60s style bare variable panels instead of realtime visualizations of the internal program state? Why do I still need to wait for a build to finish? Why is version control still such a PITA to work with? Instead everybody is jumping on the AI hype train while everything around them is crumbling into a post-apocalyptic wasteland.

      • glimshe an hour ago

        XML was going to end all other data formats.

    • pera 2 hours ago

      I still remember this one from 2016 very well:

      A Driverless Tesla Will Travel From L.A. to NYC by 2017, Says Musk

      https://www.nbcnews.com/business/autos/driverless-tesla-will...

      • bboygravity an hour ago

        Really not that far off in the grand scheme of things IMO.

        Ray Kurzweil was even more on the money in his 2006 book "The singularity is near". I remember reading some of the stuff in there that is now happening. From the top of my head he predicted 2030 for human-level AI hardware and software that could be bought for 1000 USD by anyone. I feel he's going to be very close.

    • jimsimmons 2 hours ago

      And how solving the trolley problem is the key to make it all work

  • akaike 2 hours ago

    Couldn’t it easily also take over the CEO job? Pretty sure it’s easier than producing code that works and is maintainable.

    • UmGuys 17 minutes ago

      I feel like management roles would be much more easy to automate than dev. It's hilarious how they're trying to sell these products. The only thing they couldn't do is go golf and drink half the time. They would be superior in that regard.

    • sameermanek 2 hours ago

      Given how much these CEOs are hallucinating these days while hopelessly losing money on every venture they pursue, I think AI is leaps and bounds ahead of these idiots for decision making.

    • ozgung 2 hours ago

      Good point. Maybe not for the CEO yet, but a manager without people to manage is a useless thing. So I think corporates will invent new Bullshit Jobs[1] for humans, to keep them in their organization chart.

      [1] https://en.wikipedia.org/wiki/Bullshit_Jobs

    • ben_w 2 hours ago

      I once asked a CEO, what a CEO's job involves.

      Apparently it's lots of fiduciary duties.

      As with driving cars, even if the AI is strictly better at doing these tasks than they are at writing code, mistakes aren't so easy to recover from and can destroy something unrecoverably in a 5-second attention lapse from a human overseer.

      • gf000 an hour ago

        So can software, if not more. Like, your healthcare data leaking, bank account losing your money, some legal document getting lost/wrongly issued, .. and then we didn't even talk about actual safety critical applications (which are hopefully not vibe coded) like airplanes/medical device, etc.

        • ben_w 4 minutes ago

          AI is software, so in a sense everything that can go wrong with AI must be a subset of things that can go wrong with software.

          Lots of software has a test environment. Even in live, e.g. bank account losing your money the transactions can be un-wound.

          The point is that "AI as CEO" would be in the category of "business-critical" software, and also that current approaches to AI also lack sufficient guarantees of obligation compliance or sufficient defence against failures, which in the banking example would be things like the AI deciding to save money by deleting the system capable of unwinding incorrect banking transactions.

          To the extent this kind of failure mode happens with vibe coding, it's like letting the LLM modify the unit tests to always succeed regardless of the code.

      • akaike 2 hours ago

        Well, the same goes for wrong code. One wrong line can cost millions or destroy everything completely, depending on the context. It is also not very easy to recover from.

        • ben_w an hour ago

          The two contexts where that applies is "interacts with the outside world and you deployed without tests", and "even though it only affects your own data, you don't have backups and you deployed without tests".

  • onename 2 hours ago

    Meanwhile: Nearly half of all code generated by AI found to contain security flaws - even big LLMs affected

    https://www.techradar.com/pro/nearly-half-of-all-code-genera...

    • VBprogrammer 2 hours ago

      Now do Junior Devs...

      • UmGuys 19 minutes ago

        If there are no devs, who finds the junior bugs?

      • JCM9 2 hours ago

        Are most junior devs honestly any better? In my experience no.

      • fzeroracer an hour ago

        In my experience, junior devs are actually better about this because they haven't been ground down by years of corporate bullshit and shellacking whenever they bring up security holes or issues. Some of them will be a bit over eager and miss obvious stuff but unlike LLMs you can actually train them to be better about it.

        • VBprogrammer 42 minutes ago

          Yeah, to be honest I agree. I've worked with senior people who couldn't give a crap, "it works" is the only barrier to shipping it.

          I've also worked with juniors who are technically much better than most of the senior people around them.

      • tdhz77 2 hours ago

        All devs

        • righthand 2 hours ago

          All senior devs are now junior devs, that’s why we stopped hiring juniors.

          Give me them downvotes all you want but I’m not the “senior” turning in slop Llm code.

  • anonzzzies an hour ago

    The thing is that this is probably currently indeed happening. The bizarre output of LoC by LLMs probably now eclipses whatever humans write by hand. I don't believe 90% passes the reviews though; a lot more gets written and a lot more gets discarded. My fear is more that this will change, and code will pass because less and less people know what is good or not. Not yet, but soon-ish, EVEN if the output quality does not improve.

    • bboygravity an hour ago

      90 percent definitely passes reviews.

      Source: my commits pass reviews and it's not because the reviewers don't have like 50 years of experience between them.

      It's funny how the comments on HN went from complete denial about coding with LLM's ever being possible at all like 1 year ago to "oh, it might be happening" now. I find it hard to understand how people can simultaneously be skilled (older) engineers while at the same time being completely oblivious to the concept of exponential improvement of tech in general. I guess people don't WANT to see the reality that is change and prefer to just look away and cope?

      • gf000 34 minutes ago

        Because we already are in the diminishing returns phase with relation to LLM reasoning capability - exponentials can almost always only exist for short durations.

      • anonzzzies an hour ago

        I am unsure but as a formal verification (pupil of dijkstra and his students in eindhoven and Amsterdam) person, I see this happening in a crazy waterfall and indeed people on HN are the last deniers who keep saying 'it does not work for them'. It works for the rest of the world and things are going completely ape shit now. I am happy to be old as I probably won't live through the fallout; I would say I hope we change our ways and redistribute the gained wealth here but it won't happen. And then we also do not learn from having terrible software ruining lives...

      • anal_reactor 29 minutes ago

        > I guess people don't WANT to see the reality that is change and prefer to just look away and cope?

        Yes. But frankly, it's hard to blame them. It's just a shitty social contract where you spend lots of time gaining skills and by the time you're done your field has been completely obsoleted.

        I'm a huge fan of making comparisons to how things used to work before industrial revolution, because that's what our brains evolved for. If your father was a blacksmith, you'd be a blacksmith too, and so would be your son. It was difficult to imagine a world that simply would stop needing blacksmiths. So it's completely normal that our brains spend first 20 years learning, and once we're done learning, we just apply that knowledge to everyday life, save for minor tweaks here and there.

        Then it became a requirement to keep learning your entire life because the world kept changing. This sucks, but it's not completely infeasible, so people rolled with it, especially that new positions were cozy and well-paid, and there was some gentle exit path for those who were too old to learn new things.

        But now we're in a situation where the society tells you that there's huge demand for given skill, and then your entire profession gets completely obliberated before you're even halfway done with your studies. In such environment it's simply impossible to create a realistic career plan, let alone have energy for luxuries such as raising a family or maintaining social connections. This isn't good.

        AI has objectively made our life worse, and everyone is completely clueless how the situation develops. Denying everything and hoping to wait things out might actually be just as good of an attitude like any other. Because honestly tell me - even if we assume the most optimistic scenario "AGI within five years" - how the fuck do I plan my career?

  • hakanderyal an hour ago

    For CRUD/UI heavy web apps, it can. I have probably wrote ~5% of the code I produced in the last 2 months, but I've read and verified and corrected every line of code. Output increase is substantial.

    But I'm not expecting that to hold true for linux kernel or postgres codebase or equivalent anytime soon.

    • CafeRacer an hour ago

      Im heavily using claude code right now... because i need something now. But it's a huge pile of technical debt.

      With every piece of code created i write a doc of what id need to focus on when rewriting chunks.

      AI has its usage, it makes some things faster... but spitting out more code does not necessarily make it more productive.

      I ride AI train, but i ride my cb650r same way... sometimes wondering why i am still alive.

      • hakanderyal an hour ago

        With Claude Code, you need to create the docs beforehand, not after.

        For my latest project, I have 20k+ lines of markdown docs to guide it, with great success. Some of them are generic rules, some of them describes how I code, some of them describing the codebase & features. Then I have another 17k+ that are used while coding. Plans, phases, todos, reviews etc.

        All of them are written by Claude Code also. I'm calling it "spec-driven development".

        Cursor has a much different flow, where you usually pair program with it, which I call "ai-assisted development".

        • zarzavat 27 minutes ago

          I've tried this approach but at some point you're just coding in a less precise language (English), and with more duplication.

          My current approach is to stub out the architecture I want, write the type definitions I want in full, and ask Claude to fill in the blanks. It's good at implementing an interface but bad at devising the interface on its own.

          • hakanderyal 3 minutes ago

            That’s more or less what I do, but in an expanded way.

            It’s not coding in English, but giving Claude memory about how the code base works, how it’s structured, what features available, why things are created etc.

            This is how we work too, we have all these info in our head when building. But every CC session starts a new with blank slate. These documents helps getting the session “warmed up” enough to get quality output with lesser tech debt.

  • edg5000 an hour ago

    For me this appears to be true actually, these days. It struggles more with large and terrible codebases though, because it has to ingest so much garbage code in order not to break anything, wheras with nicely strucured code it can do great work on a module level, as long as you direct it well.

  • fleebee 2 hours ago

    As a note to future historians, this sounded just as crazy back in March.

  • poniko an hour ago

    Did some bigger admin tool to do search and matching datapoints etc and decided to only use claude code. First version was wow awesome this saved me so much time .. now 2 weeks later putting back code that was deleted several times, removing 3 copy's of the same code, way to complicated sqls, verbose code and looking how its mixing htmx with some wird own JavaScript. It really looks as a junior developer solutions, so I'm done using it for more then boiler plate things..

    I will come back in 3-6 months and hope its better to understand its own limitations.

    I'm guessing it would be less then a week for me to write it on my own.

    The biggest issue is that I've now seen the shit it creates so I have zero trust in the code I now have from a security and stability standpoint. I know many have better experiences then mine.

    • verdverm an hour ago

      Have you tried agent instructions? They have been helping with these issues

  • zarzavat an hour ago

    "AI" has been writing 90% of my code, on a per-character basis, since IntelliSense came out.

    The question is not how many characters can a computer spit out, but whether you need a human in the loop or not. That's not going to change in 6 months. In a few years, perhaps.

  • urquhartfe 2 hours ago

    Well, he still has a month I guess

  • verdverm an hour ago

    I just did a week with AI coding and it did about 80%

    That being said, I had to be very detailed and meticulous because they go off the rails and make poor decisions

    Just means I type and search a lot less, and can get things out of my head and into computers much faster

  • heisenbit an hour ago

    I'm fairly sure in 3-6 months 90% of CEO pronouncements are prepared with AI.

  • lawgimenez an hour ago

    This might be true, but in my experience the code generated by AI are mostly a bunch of bad practices.

    Project managers be like sending me AI code snippets like they know shit.

  • UmGuys 21 minutes ago

    If the general public believes this nonsense, I understand why there's so much AI hype. He's just selling his product like Musk, right? No one with a surface level of context could believe AI will replace all developers by March 2026 and soon after all human employees.

  • snowstormsun an hour ago
  • password321 2 hours ago

    So when will we get AI automating AI development?

    • ben_w an hour ago

      That started years ago.

      Exactly when depends what you count, but part of the path leading to ChatGPT was getting a bunch of dumb systems to work together to train a smarter system: https://arxiv.org/pdf/1909.08593 fig 1, and also section 4.4 for what can go wrong

      • password321 an hour ago

        It wasn't a serious comment but my point was about seeing Anthropic/OpenAI replace its own software engineers.

  • r0ckarong an hour ago

    Nobody said anything about working software.

  • tobyhinloopen 2 hours ago

    I use AI a lot while coding and even I don't get close to 90%.

  • dachworker 2 hours ago

    I mean, trivially true, if you consider that AI-enhanced programming requires 1000% more code written, deleted and rewritten, countless times, by AI itself in it's feeble attempt to "reason" about the problem.

  • JCM9 2 hours ago

    Probably a bit ambitious, but I do see a future not too far off where junior and mid-tier developer roles basically disappear.

    • BoingBoomTschak 2 hours ago

      Thing is that experiencde programmers must start at "junior" like any other...

      • JCM9 2 hours ago

        You’re not wrong, but there’s a glut of junior developer talent at the moment that there’s no shortage of folks to take on those most senior roles. This this is a “tomorrow” problem that (for better or worse) will get ignored for now

      • thom an hour ago

        I suppose the question is whether the only way to educate good programmers is for them to spend time as a (possibly net-negative) apprentice in an enterprise job. Perhaps under the watchful and patient gaze of future AIs, they're able to blossom earlier, perfect their skills on open source, and enter into the workforce ready not to cause chaos. Personally I'd love the alternative version of the future where LLMs do all the bullshit (meetings, alignment, planning) and then seniors feel far less stress and pressure and rejoice in being able to spend time raising up their fellow humans.

        Not sure either outcome is likely though.

  • kosolam an hour ago

    “Writing the code” is not the same as programming. The ai WRITES most of the code for me, but I’m still doing 95% of the PROGRAMMING.

    • ethan_smith an hour ago

      Exactly. The AI writes the implementation details while I'm still doing the actual programming: defining problems, designing architectures, ensuring correctness, and making the difficult tradeoffs that require domain knowledge and judgment.

      • verdverm an hour ago

        I wonder how much of the sessions will be used to train up agents that can do more of the programming. I'm certainly providing lots of feedback so they don't mess up so often in the future

        Someone also said extending this pattern into the future is probably why computers in Star Trek mess up

    • bboygravity an hour ago

      Writing the code? How about reading it and explaining it?

      Working on undocumented legacy code (read: probably the majority of all embedded code in the world) is now like 100x faster. No need to plow through a massive codebase and trying to grasp expired toolchains for weeks/months anymore.

      How is that not part of programming?

  • z3c0 2 hours ago

    Amodei's work history indicates that his background as a software developer is a single part-time job that he held for a year-and-a-half after college. As far as I'm concerned, he wouldn't even make it as a junior on my team. I'm not inclined to believe anything he says about what it takes to write production-ready code.

  • newsclues 2 hours ago

    Has an AI tool been used to develop a new programming language (perhaps that is better for AI than existing programming languages?

    Or another way, if Rust didn't exist, could an AI create Rust?

    • wordofx 2 hours ago

      I would have to go searching and can’t be bothered. But there was a guy on Twitter earlier this year who said he had been wanting to create his own programming language for years and with AI he’s been doing it finally. I cannot for the life of me remember what it was called.

  • crinkly 2 hours ago

    I remember calling bullshit on this well over a year ago. I was shot down by my colleagues and management. I will remind them about this on Monday :)

    • verdverm an hour ago

      I had my doubts, then I spent time getting to know how to work with and instruct them, game changer that you don't want to write off in arrogance

  • ath3nd an hour ago

    GitHub CEO delivers stark message to developers: "Embrace AI or get out." (then proceeds to get out himself, 7 days later)

    https://www.businessinsider.com/github-ceo-developers-embrac...

    https://www.theverge.com/news/757461/microsoft-github-thomas...

    This industry is full of snake oil salesmen and false promises, it's reliant on hype and an army of fanboys trained on twitter to fan the flames and create the impression of a revolution that isn't there and might never be there, just to prop the valuations of the likes of OpenAI and Anthropic.

    It always follows the same playbook:

    - The guru type e.g muskie, sama, adam neumann, elizabeth holmes, SBF, says something on twitter: "XYZ is the future"

    - Army of same-thinkers lurkers start getting excited about it.

    - VC fund bros who are also forming their opinion on vibes stalking twitter, also get excited about it.

    - Big money poured in XYZ thing.

    - Guru keeps claiming that XYZ version 2.0 is just behind the corner (AGI/Superintelligence, Autonomous Self Driving Cars, the Boring Tunnel, cure for whatever) projecting insane and out of touch with reality timelines.

    - XYZ thing either becomes the future (not likely), or gets abandoned (remember when buying pixelized pics of apes on the internet was an "investment"?) and the collective amnesia of the fanbase doesn't keep the guru accountable. It's easy to make hundreds of claims if getting it right once, even by chance, makes you some kind of a prophet and all your wrong claims of the past are dismissed due to your fanbase's dog-like adoration for you.

    - Cycle repeats.

    The LLM craze reminds me of the NFT, web3, crypto grift all over again, but this time with a product that at least has some value.

    Tom Renner wrote an excellent article on the "LLM inevitabilism" https://tomrenner.com/posts/llm-inevitabilism/

  • mojuba an hour ago

    There was this wave of outsourcing of engineering jobs to countries with cheaper labor back in the 1990s and 2000s, all at the expensive of quality, accompanied by mass layoffs in the western countries of course. The industry was testing the limits of how much lower the bars of quality can go until customers stop paying for software, and as it turned out, customers could tolerate a lot.

    Now, AI is the new outsourcing it seems. Let's lower the bars of quality, see how much crappiness and glitchiness can be tolerated, freeze it right there as the new "industry standard" and move on to the next cost cutting thing.