Settings

Theme

AI came for the coders. Understanding why tells us who's next

robertpeake.com

59 points by speckx · 68 comments

Reader

10 threads
mattkenefick

I think it most publicly came for coders, but has been quietly taking over everyone else without them knowing it.

  • a34729t

    90+% of lawyers, per my lawyer friends, are at risk. High end, safe for a while.

    • edgarvaldes

      Isn't the legal profession a "protected" one, that is, a practice that requires a certified professional?

      • dotancohen

        Yes, but now each certified professional can handle an order of magnitude more clients. Either the client base grows, or 90% of the attorneys will be out of work.

        Worst case scenario: a little bit of both.

        • atmavatar

          I'll inject a little glimmer of hope: perhaps it will aid perennially overworked public defenders in handling their case loads, helping to ease one of the greater injustices against those without means to hire their own private lawyers.

          • throwaway63467

            It will just create more work as people will sue more, already now courts get flooded with AI generated slop, AI even eagerly recommends this and helps people to submit docs as well.

            I see it in my job as well in privacy, I can review more stuff so in theory it saves me time but I get 5-10 more cases to work on so in the end it’s more work.

        • karmakurtisaani

          Maybe legal representation becomes more affordable, as it should be.

        • wooger

          All the trainee / junior lawyers who spend their first N years reviewing casefiles and doing donkey work are no longer needed (unless things aren't digitised).

          An AI can write a letter in legal wording threatening stuff for free.

      • sl0wbrainRot

        Having been through the civil court process this year, it can absolutely be automated.

        Have all parties provide access to bank and email, whatever accounts court wants data from...

        AI summarizes it. Judiciary can skip hearings and review.

        They don't really take the oral side of things seriously anymore. 100% felt like an empty formality.

        Constitutions usually only define Executive, Judiciary, Legislative type roles. The protections you mean are just laws that can change

        • kvdveer

          I think you're missing the purpose of all these processes. They're not there to ensure a fair trail (that could be done far more efficiently), They're there to prove beyond reasonable doubts that the process is fair to all those involved.

          This is not a role a black box AI can fill right now. Not because of any missing capability, but because of missing public trust.

          • sl0wbrainRot

            I think that's the verbatim copy paste of the philosophy but not tethered to reality

            At least in my experience

            The case I was a part was very much cut and dry law, based on finances and contracts; no doubts as the digital records were right there

            Lawyers in the case just parroted almost verbatim stuff you would hear off Law & Order. Outside court they coached us to feign ignorance. The judges know this as they were lawyers before being judges

            Some family court stuff might need human intervention but that should be medical experts and such who submit their findings straight to court.

            Lawyers goal is win not "justice". They will manipulate knowledge of available records and evidence to fake reasonable doubt and they don't always get caught.

      • dist-epoch

        Laws change all the time.

        Taxi driver used to be a "protected" profession.

        • edgarvaldes

          It seems to me that the branches of government (the ones capable of enacting a change of this magnitude) are more closely intertwined with law firms than with taxi drivers, at least at the federal level. The incentives just aren't there. I suppose time will tell.

          • dist-epoch

            The top partners remain, all the rest are fired and replaced by AIs. Seems like a win to me.

      • matheusmoreira

        Yes, and those credentials are literally the only things stopping AI from establishing a beachhead inside their field and advancing.

    • matheusmoreira

      Doctors too.

      • ilovecake1984

        not so sure about that. Being a doctor is a physical thing.

        • wooger

          I have very little faith in my recent GP visits, to the point that I'd trust a random intelligent person with AI to diagnose me better.

          I've felt for a long time that doctors are vastly overtrained for what most of them are doing 99.99% of the time, and some more lightly trained people with AI assistance and escalation to a proper doctor would be perfectly fine. And much cheaper and more numerous.

          Meanwhile specialist doctors should be more specialised and not waste their time in generalist training for 5+ years.

          This is all partly informed by UK medical training being ridiculously long, not enough places, now insanely expensive, while working conditions once qualified are dogshit and meanwhile there aren't nearly enough doctors to fulfil NHS requirements. The lack of training places could be easily fixed by a competent government you'd have thought, but still seems a waste of time to spend 8 years to qualify as a GP.

        • 10111001

          > not so sure about that. Being a doctor is a physical thing.

          In the dystopia we live in, someone's going to replace that with whatever scans and cheap and available + AI, and set up the incentives so people are forced to use that instead of go to a real doctor. They'll do that even if the outcomes are poorer, justified with some apples-to-oranges comparison.

        • blacksmith_tb

          Yes and no, it's easy to picture a loop like "take patient's reported symptoms, identify most three likely diagnoses, order tests to confirm / disprove each." A human doctor might listen to your lungs, but I can also see a cheaper-to-pay medical assistant taking recordings of your heart, inhale/exhale etc. to feed in as inputs.

          I would say the only non-obvious aspect is liability, currently doctors in the US carry large insurance policies for malpractice suits; if your treatment was handled by an AI system, who'd be liable for a misdiagnosis?

    • dyauspitr

      Law will be fine because the way things are legally set up even if AIs do 100% of the work the lawyer will still have to rubberstamp it. If anything, their lives are gonna get really cushy.

      • hagbard_c

        > If anything, their lives are gonna get really cushy.

        That is, until someone decides he can do a Ryanair or IKEA on the sector and starts offering law services at 1/5 of the cost for the 90% of the cases where such services are sufficient. Suddenly that new BMW every other year seems to be less of a given.

        • solyin

          RocketLawyer's been around since before ChatGPT. Depending on what you need, the comparison to IKEA is apt. Great for generic things that everyone needs/wants; less great for specialized custom stuff. Throw in some AI and it's not clear the timeline.

          • hagbard_c

            Interesting you mention RocketLawyer given that I recently received this message:

            De Nederlandse Rocket Lawyer website zal niet meer beschikbaar zijn na 31 oktober 2026. Zorg er dus voor dat je alle documenten afmaakt en downloadt voor die datum.

            ...which translates to:

            The Dutch Rocket Lawyer website will no longer be available after October the 31st 2026. Make sure you finish and download all documents before that date.

            I don't know if this is a prelude to an exit from the Dutch/European market but something is changing there, for sure.

  • hoomanyyy

    yes i think we should support this kind of startups more

datadrivenangel

AI for coders is somewhat analogous to a backhoe or automated welding machine. An operator can do a lot more work than before, but setting it up to do that work correctly and safely still takes skill.

This however will put salary pressure on the people who were simply digging holes or welding beads without being as involved in the planning and design work. Time to remember that we can't forget the business part!

  • spiderice

    > but setting it up to do that work correctly and safely still takes skill

    For now. Not sure why you think that would last though. Obviously the big labs are going to make it easy to use AI both safely and correctly. It's already significantly easier to do that today than a year ago.

    • geraneum

      > Not sure why you think that would last though.

      The failure modes of the LLM are still the same. Nothing has fundamentally changed, despite the UI and the harness showing significant improvements and outcomes.

      • wolttam

        I almost never hit the failure modes of LLMs anymore, they have been getting much more reliable year over year.

        More and more it would appear the failure modes are a training data issue

    • dotancohen

      I completely disagree. The AI tools I've used from the big shops are happy to code visible features, but need real restraint to keep things in order.

      I no longer really check the code, but I won't let any of them touch the database schemas. I modify the schema, the machine codes around my database. Same policy with APIs, even private ones.

      • equinumerous

        > but need real restraint to keep things in order.

        I second this. Even OpenAI's flagship models... the code I write by hand has much less entropy and is easier to follow than the equivalent AI-generated code. I think part of the software engineering skill now is knowing which pieces you can delegate to AI and which still need human input. I know many people who just YOLO it all and have AI do everything... but, this can hardly make for a stable system in my opinion...

        And, I'll be happy to be proven wrong, because it will mean a significant amount of work is off of my shoulders. But I think we aren't quite there yet in terms of having AI code that lowers the complexity instead of raising it in a codebase.

        • bentobean

          I hear you, and I agree with you. But the response to this by leaders at my company would be:

          > Why are you looking at the code? Complexity? What do you care? Let the agent handle that.

    • docshock

      Your trust in the big labs may be misplaced. What do they rent you? A black box that cannot be inspected.

      If you actually start looking inspecting what the boxes do, it is a scary place. The Chinese models are open source open weights and are trained on lies. This much can be proven. But the frontier labs have made it impossible to determine the provenance of their models. They have been fed with copyrighted works, but good luck proving that your copyrighted work lives in their weights.

      In short, the frontier labs are operating opaquely, and that's very good reason not to trust them.

  • dan_grec

    > An operator can do a lot more work than before, but setting it up to do that work correctly and safely still takes skill.

    An architect friend just released a full game on Steam, and it’s making money.

    He has never written code is his life, has no idea what a database schema is, or exceptions or even a shell script. Never heard of git, ssh or really anything else. AI wrote everything, even the textures, sound effects, music, etc.

    Coding is done.

  • matheusmoreira

    > Time to remember that we can't forget the business part!

    There is no reason why AIs wouldn't be able to replace CEOs. It'd be a crime against humanity for us to stop just short of replacing executives.

    • docshock

      There are companies that want to do exactly this and set up an entire virtual c-suite with zero accountability. A company so constituted is patently not a person, but this is not a matter of settled law.

      • bentobean

        Corporations as a service. Interesting.

        You know, it occurs to me… it might be interesting (and funny) to setup some kind of LLM hive mind consisting of agents assigned various asinine corporate roles (“you are hr bot”, “you are cto”, etc…). Feed random problems into it and see what it attempts to do.

    • 10111001

      > There is no reason why AIs wouldn't be able to replace CEOs. It'd be a crime against humanity for us to stop just short of replacing executives.

      They'll go, but they'll be the last ones.

      The only thing that won't ever be automated is ownership of property, which is the only thing that's sacred to the wealthy.

    • mxfactorial

      zero commits from u on gradienterp.cloud

drdaeman

> there will never be a compiler that can check for bugs in your poem.

Why not? There always were and will be artists who would want to crack their art open, analyze it, understand how it works, and synthesize new art intently, engineering it. And no amount of sacralization of that air of mystery around art will stop that from happening. I’m absolutely sure there will be machines that analyze poetry, model readers/listeners (contemporary or predicted future audiences), and find “bugs” as mismatches between the intended and modeled effects.

And the art won’t be any tiniest bit less enjoyable. Well, unless present-day adtech will survive and abuse it all.

  • gAI

    "In the eyes of those who anxiously seek perfection, a work is never truly completed—a word that for them has no sense—but abandoned; and this abandonment, of the book to the fire or to the public, whether due to weariness or to a need to deliver it for publication, is a sort of accident, comparable to the letting-go of an idea that has become so tiring or annoying that one has lost all interest in it." - Paul Valéry

    "The repetition of an experiment results in the accumulation of evidence, which either proves or disproves the validity of a hypothesis. The repetition of a work of art does not prove anything; it's just a copy." - Patrick Dunn

    • drdaeman

      "Is it possible that software is not like anything else, that it is meant to be discarded: that the whole point is to see it as a soap bubble?" - Alan Perlis

      "Any headline that ends in a question mark can be answered by the word no." - Ian Betteridge

      P.S. Patrick Dunn, really?

danpalmer

Software Engineering is objectively measurable, but needs a different objective measurement (test, benchmark, etc) for each part, feature, behaviour, etc. This is part of why I don't think it's going away as a discipline, although I agree it'll change very significantly.

The areas hit hardest are going to be those where one person can specify the objective once, and then value can be delivered ongoing from that. Data labelling, categorisation, maybe things like QA. Arguably things like autonomous driving.

  • GrumpyYoungMan

    LLMs can only do QA in the sense that a compiler can find bugs. Yes, it is very effective for various categories of mechanical mistakes, more of them than a human can keep track of and faster than a human would. But LLMs mostly are not going to be able to detect the sorts of architectural, conceptual or requirements mistakes that a good QA person can.

jimbobbam

Great article. I agree with the sediment entirely.

Still though, we have major problems that need solving. Broken infrastructure, rising tides, deforestation, loss of bio diversity. It will take people to stop looking at the box, and start looking at the problems that need fixing.

itaintpretty

> Yet, as I have said before, there will never be a compiler that can check for bugs in your poem.

I’m quite tired of the code is verifiable bullshit. Of course, it compiles or not, but that’s not the same as did you produce a coherent system, with proper boundaries? Did you decompose the problem properly?

People seem to think that is somehow the easy part. But that is not verifiable at all. Proper abstractions, division of responsibilities etc is not easy and requires non-trivial amounts of knowledge on human psychology.

The clankers are still getting better at it and that stuff isn’t verifiable. Or if it is, pretty damn near everything is. How is law different, finance, etc?

They still aren’t perfect but the fact they are improving on something so intangible should scare the S out of us. Nothing is safe.

  • zahlman

    > People seem to think that is somehow the easy part. But that is not verifiable at all. Proper abstractions, division of responsibilities etc is not easy and requires non-trivial amounts of knowledge on human psychology.

    They seem to believe that these things are not necessary as long as redoing the verifiable aspects is cheap/fast enough.

snailcoder

The profession so good at their job they automated themselves out of it.

moishe_syzslak

I don't know how good this poetry major is at coding, but his writing sucks.

So much of what has been automated most of us lament, and so much of what has not been automated we begrudge.

Wut?

  • black_rabbit_

    Something like:

    > Most of us lament the things that have been automated, and begrudge the things which have not been automated.

teleforce

>I used to joke that I got my degree in poetry instead of computer engineering so that, in case this computing fad went south, I would have something to fall back on. Thanks to AI, it’s barely a joke anymore.

Ironically the author got his degree in computer engineering (CE), and work as software engineer (SWE). If his current SWE carwr went south, he should revert to his CE career. Last time I checked Nvidia, a classic computer engineering company, is the world's most valuable company. Together with Micron, another classic computer engineering company, they form an impressive 34% of US earning boom [1].

By his own article projection poet probably be next so called "victim" of AI.

>Capitalism keeps people answerable with an existential threat: if you don’t work, you don’t eat.

Perhaps it's the probably the problem with rampant capitalism, and AI is just happened to be the most recent scapegoat.

The article is really negative in its narrative and, me and many others have addressed this negative AI naratives many time at HN [2]. It becoming boring and tiring, and this one is probably one too many.

[1] Nvidia and Micron Make Up 34% of America's Earnings Boom (Oct 2026, 47 comments):

https://news.ycombinator.com/item?id=50236073

[2] There's never been a better time to go into math (by: 3blue1brown) [video] (comments section, Oct 2026, 824 comments):

https://news.ycombinator.com/item?id=50148916

fooker

And yet, there are more coders than ever before.

  • wooger

    *In developing countries, soon to be let go.

    The number of coders can fall sharply too, not hiring any grads, and especially not hiring any immigrant visas or outsourcing stuff to India will drop that number a tremendous amount.

  • deflator

    Yeah, I keep hearing this but my experience is the opposite and more in-line with Jevon's paradox

    • dron57

      Jevons paradox would in fact imply that there would be more coders since the demand for coding would have skyrocketed. What do you mean your experience is the opposite? Are there fewer jobs now?

      • deflator

        Apologies. I was agreeing with the parent.

        I should have said "Yeah, I keep hearing this but my experience is also the opposite and more in-line with Jevon's paradox"

      • pixl97

        That the pay of coders could fall so far that it's not a viable way to make a living.

        • dotancohen

          Coding is no longer a viable profession for me. I admit that.

          Selling software, however, that's still going strong. I just don't code it anymore. The machine does that.

          • soon_to_be_sb

            Yes, Right now and going forward, most software shops will be one LLM and a hundred marketing guys...

            • docshock

              The technical deficit of such an operation would be staggering, and the curtain would soon fall on the fraud. However, our legal system is not prepared to find this activity fraudulent.

      • Supermancho

        > The paradox is an economic theory stating that technological progress which increases the efficiency of resource use leads to a rise, rather than a fall, in total consumption of that resource

        I understood it differently. The paradox doesn't mean more coders any more than Combines necessitated more field workers. More compute, releases and more code changes, yes. I don't believe an increase in demand for developers has manifested as the result of AI tooling becoming widely used.

      • saghm

        Yeah, this is a very confusing argument. I don't understand how someone could interpret Jevon's paradox the way they did. It's like saying that more traffic means there are fewer cars on the road.

  • add-sub-mul-div

    That doesn't mean what you're implying it means if there's an influx of shovelware generators.

  • alephnerd

    A lot of people are basically using AI as an excuse for their skill obsolesce. It's the same people on HN who would pooh-pooh cloud, JS, k8s, and other skills that are increasingly critical for tech.

Keyboard Shortcuts

j
Next item
k
Previous item
o / Enter
Open selected item
?
Show this help
Esc
Close modal / clear selection