I open this discussion thread and literally the first two top-level comments I see contradict each other:
> I find that productivity follows a Pareto distribution (80:20 rule) and that AI is a sharper tool in that it enables the already productive to be proportionally more productive and this effect increases as the intelligence of AI increases. So 80:20 becomes a 90:10, 95:5, 99:1 etc.
> Big caveat to the productivity claim is that it’s concentrated in lesser experienced engineers and vanishes or goes negative with highly experienced engineers.... LLMs move your baseline towards the mean. If you are below average it improves and if you are above it hinders.
While I disagree with the counter position I do believe it to be an honest account of their experience. I think the counter intuitive aspect is that marginal cost of software is practically zero so in an efficient market the consumption of it will be dominated by the best software written by senior devs with AI, even if the production of software is dominated by junior devs with AI. I think we are in a transitory phase where still employed juniors can be very productive shortly before becoming unemployed at which point that productivity drops to zero. If junior devs are productively producing software that then goes unused then how productive were they really. But I understand that the latter effect may not have been picked up in the general stats. Anecdotally again, I fired 3 juniors and went solo and I know of many others who have done something similar.
Another aspect is that once the easier work is complete then the remaining residual is the difficult work, and this changes the relative productivity to the point where a junior developer may not be able to make any further progress at all.
Overly bureaucratic companies tend to generate easier work for themselves to the point that this work dominates their total workload. If we are to measure productivity by opening and closing Jira tickets then I would expect the junior developers to get a much bigger ‘productivity’ benefit from AI than senior developers.
> If junior devs are productively producing software that then goes unused then how productive were they really
Most ways of measuring productivity across an economy (in macro sense) takes into account only the transfer of capital. Meaning someone paying for something. So in this case it will be measurably zeo in most macro-economics studies.
> it enables the already productive to be proportionally more productive
I finally got a real feature to develop and started using Codex a couple weeks ago. Before that, I was just using ChatGPT for small things. For the most part, I was doing a lot of monkey-patching and deep troubleshooting that I didn't think Codex would be good at.
I was wrong. I had no idea how smart these tools are.
I now realize, as the tech lead on a large project with me, 2 front-end developers, a full-stack-ish developer, a tester, two product people, and a PM - that I could do this entire project more efficiently with just me and the tester.
I'm doing SQL for the first time in 20 years, which would have slowed me down in the past, but not now. It used to take me an hour to program a new feature and two hours to get it to look right, if I had to do the CSS myself. Now I can do that w/o anyone's help. Writing tests used to take as long as writing the code. Maybe it still does, but that's an hour instead of a day. And it's so much less painful. There's no satisfaction of the aha moment when writing tests. It's just a slog.
We're a very small shop that usually only has 1-2 devs on a project. So we're not set up to do big projects, we aren't doing agile, and it's just a mess. We waste probably 80% of our effort just communicating.
The product team was needed/useful in the beginning to define the broad scope of the project and lay down the UI/UX patterns. But now it's like pulling teeth to get them interested in our internal client's ever-evolving wishlist. I can handle that, and always come to the product team with very specific asks when I need UI design help. Just give me the tester, who's so thorough it's annoying, to catch the weird edge cases I hadn't thought of.
I'm not saying I'm God's gift at any of these things, just that I'm reasonably competent at them. It's kind of terrifying what a seasoned full-stack dev with a decent feel for UX and the ability to manage stakeholders can do with this.
It's not so contradictory if you peel back a couple layers
1. For junior engineers a 50% increase might be less than a 10% increase for senior engineers. So the % comments are consistent with the pareto argument. Diminishing returns _can produce_ a pareto tradeoff.
2. Saying gains are "concentrated" to junior engineers can also be a function of the work assigned to junior engineers. A senior engineer might get 1000% boost over jr doing the same jr work. The problem is the reverse - the hard things are concentrated around the oldest engineers, and now "hard" can also mean "not easily claude'd"
Yes, I’m a senior engineer at a 3-engineer startup, and there isn’t “junior-level” or “senior-level” work here. It’s all just work. Claude allows me to tear through the kind of gruntwork I’ve been doing ad nauseam for the last 15 years.
If my tasks were things like squeezing 3% more efficiency out of an algorithm, that might be different. Although, even then, my PM (!) used Claude to boost the efficiency of one of our more complex Postgres queries by up to 5x, in ways I honestly would not thought of doing. It was quite humbling. A former engineer here used pgMustard on the same query a couple years ago and claimed it was as fast as it could get.
I can’t speak for the kind of work done in the higher echelons of big tech, but for us, there isn’t much that Claude can’t help with.
Yeah those single points of failure/bottlenecks are prefect focal points for people with floating responsibilities and a good understanding of how to use claude.
The system-sweeping features I still would ask to review, but you can't argue with a DB query just getting faster.
> Although, even then, my PM (!) used Claude to boost the efficiency of one of our more complex Postgres queries by up to 5x
Would you be kind to describe what test set-up do you have that a PM felt safe enough to deploy the changed query without the fear they'll break something?
That is not the most absurd, the most absurd is people here thinking faster ( and yes I avoid the word better ) code generator engines will solve humanity problems.
Terrence Tao is an exception in many regards. I’m less concerned if it helps him, and more interested in how it affects the average person, since the vast majority of users will be average.
Tao can use LLMs on his own terms, on his conditions. Under those circumstances, it's an absolutely fantastic tool.
In an enterprise, everyone gets to use LLMs. Everyone can create slop. In an enterprise environment, this can lead to weird situations where the senior has to put in extra efforts to contain the slop of others... so I can kind of see where the "hinders seniors" is coming from.
I don’t think it’s even “seniors” containing slop of others.
I think it’s those who still care about quality (which seems to be decreasing in the industry overall) trying to keep the whole thing running, in a sea of people who think going really really fast and not understanding their code is fine because LLMs can make sense of it and clean it up later.
Just look at the condition of open source contributions recently
A challenge with this kind of study is that coding agents (Claude Code, OpenAI Codex) only started working really well in late November, which for most people meant early January due to the December break.
General agents (OpenClaw, Anthropic Copilot, ChatGPT "Work") started working even later than that.
This category of software may have a much more meaningful impact on work than the mostly-chat systems we were using from 2022-2025.
Studies that mainly focus on 2022 to end of 2025 might be missing out on a material uptick in capabilities.
Anecdotally, I (along with the rest of my team) got laid off from a big tech company at the end of January. The stated reason was "AI", although I think as we all know the real reason was "we overhired in 2022". I managed to get a job by April without putting in a lot of work, though (mostly just listened to recruiters until something hit my fancy). My work experience is good, but I don't know if it's so good that I would be immune to market effects. The company that hired me is still hiring other software developers, and I was told that it took them a long time to find someone with my qualifications. (We use Claude, although so far I haven't seen a lot of AI psychosis like I have in other places.. that was also a big filter in my job search). Long story short, I think narrative of job loss is way overblown. It's more doom trolling from anthropic and openAI and their mouth pieces as far as I can tell.
It's so early game, is the real problem. My experience so far, has been that it's barely a junior dev. I've met so many in my career that think reading stackoverflow, or watching a youtube video mysteriously makes them an expert. AI reminds me of this sort.
Well anyone can use (prior to AI) a simple linter and learning to code isn't that big a deal. It's learning the pitfalls, the traps, that's the issue. And so far Opus just seems to fall into them again and again. I guess the best way to put it, is that it's not an architect. I sees no big picture, and that's not really a surprise with (compared to a human) an incredibly small context window. When I'm on a project, or working with a codebase, I often have years of "context window". And I have a career of "don't do this" context window.
So what I wonder is, will this be resolved? Will that awareness of larger scope be solved? If that happens, we'll be in another ballpark of competency.
Some companies have massive codebases. Are these companies slowly gaining rot in those codebases, a swiss cheese effect, which eventually will result in collapse? Because I've worked where a bad hire had just this effect over time. And what I worry about isn't using Claude to speed one up, it's the DEV that uses Claude and just "meh" and submits because it passes regression + other tests, and then a manager or code reviewer uses Claude and "meh" because it's a pass too.
> My experience so far, has been that it's barely a junior dev.
At least on greenfield projects, Fable is more like having an endless succession of fly-by senior devs who lock themselves in an office for a week, and who come back with very reasonable code that I then need to maintain somehow.
For longer-term maintenance, I am actually slowly warming to Sonnet 4.5-era models (so October 2025, right before the Opus revolution). They need to be given clear instructions and watched carefully. But since they force a human to stay in the loop, you don't have the institutional knowledge loss I see in some Opus projects, or the code that was one-shot with no human interaction at all that's a constant temptation in Fable projects.
And yeah, the bit rot is painfully real once the humans step back too far. I've been dealing with a compelling prototype that someone built, and that stakeholders love (for good reasons). But it had to be put on a tech debt repayment plan for a couple of months.
> Are these companies slowly gaining rot in those codebases?
Yes!
And that "yes" holds regardless of whether they use AI or not.
Let's not pretend tech debt accumulation is somehow an AI problem. Some of the world's biggest companies routinely ship code that reeks of years of rot and decay.
It's safe to say that most people romanticize their past experience. Anyone's "golden age" is someone else's "end times".
Software has always sucked [1]. "Software crisis" was coined in 1968. There has been no period of time since that it got any better. Trauma just doesn't carry the same way that triumphs do.
Perhaps people do not actually learn from their mistakes after all.
> Let's not pretend tech debt accumulation is somehow an AI problem. Some of the world's biggest companies routinely ship code that reeks of years of rot and decay.
While that is true, there's a large difference you are ignoring: a company can survive this type of rot indefinitely if it takes place at a certain rate. Their customers get time to acclimatise , the processes take into account the breakage, and adapt to compensate, etc.
If the AI boosters are to be believed (2x, or 10x or even, in some cases, devs claiming 100x productivity), stuffing 2 decades of rot into the next two years could prove fatal.
Our production incidents have dramatically increased over last 2 years after management pushed AI. In 2023, we never failed our SLA. Now, we continuously keep failing it due to some stupid outage. There was no real change in architecture or strategy - only the adoption of AI.
There have already been several high-profile AI-root-cause incidents at many companies this year. I guess we can look forward to even more deterioration in the next 2 years.
I have seen a bad hire having instant impact on team and after person was let go, team was still getting back on track to pre-hire productivity for another 3 months.
We know all managers that say "stack doesn't matter, we can swap devs". We had C/C++ dev hired to do web dev with C# and Typescript.
Guy was utterly incompatible even if he was smart but management wanted someone who is local and can come to the office and that was this guy selling point.
I get the point having read much the same from Tesla (and fans) regarding self driving cars that still haven't done half the things that Musk said was just around the corner pending regulators a decade ago and repeatedly since then.
And myself I keep making comparisons between AI and the progress in 90s video games where every minor improvement got called "photo realistic" and then forgotten with the next game engine: https://archive.org/details/nextgen-issue-26
So I'm not gonna say "this is it" when the software quality really matters, and I absolutely won't speak to progress (or lack of it) outside of software.
But I will say "you can look around and easily see small businesses using AI to generate posters, quite a lot of small business software and websites are in the same category: the mistakes are real but increasingly don't matter".
>the mistakes are real but increasingly don't matter...
I think it would start to matter once again. People will get fed up of AI posters and art. I think they already are...and once some threshold is crossed, the business won't dare to use AI generated assets/designs.
Turns out humans are much better at recognizing patterns in stuff that is generated ONLY using patterns from human generated content.
> People will get fed up of AI posters and art. I think they already are
Agreed, but will this look like a meme/fashion cycle? If so, re-prompt each year with a different look. Yes, there are still issues here, a friend found an image he was amazed was AI generated, but to me it was obviously so, so I showed him a screenshot of ChatGPT making something just it and included my prompt:
create image: hand drawing of cute springer spaniel puppy looking sideways, various geometric shapes drawn in layer behind and in front of the puppy, all done in style of 7 year old using crayons with mediocre colouring-in skills
As I said to them:
yeah, the line thickness feels AI, to me, the bad colouring-in scribbles feel like just the art style it was propmpted with
it's like: it gets the big picture of the composition, and it knows how to colour in badly, but it doesn't know how to draw a dog as badly as the colouring in
> Turns out humans are much better at recognizing patterns in stuff that is generated ONLY using patterns from human generated content.
We're better at recognising patterns full stop. All biological brains are, and needed to be better than the current state of the art in machine learning because if a living organism was as poor at learning patterns as the SotA in machine learning, the organism would starve to death before being able to pick up anything and eat it.
AI also has a second disadvantage, because there are so few models: the laziest of ChatGPT "thinkpiece" blog posts being everywhere is hard to miss, and 5000 fake bloggers all prompting the same model with "find biggest news story of today and write a blog post about it in a way that maximises my ad revenue" will get 5000 almost identical posts. This will remain true while each instance of the most commonly used AI fail to talk to each other in a way that at least mimics them collectively getting bored with writing the same thing 5000 times, it does not depend on e.g. quality.
I digress with your last quote, I feel like some people who are aware of AI are starting to develop a quasi allergic reaction to slop, and while the mistakes might not matter for most of the population, for others it does and will take notice.
I’m probably illustrating your point but as a FSD fan it really got ”good enough” recently with version 14. The tipping point was suddenly, much more often than not, it can drive end to end from start (my garage) to finish (parked at destination) with no interventions. I can text and watch videos on my phone and as long as I glance up once a minute, it doesn’t complain.
Handling highway driving with lane changes was great when it got there years ago, but just in the last year or so it has gone from a nice to have to “from now on I will never buy a car that can’t do this”.
AI has hit some milestones for replacing work as well. There’s still many more to go and maybe some of them will never get hit (much like I don’t think a coast to coast drive with zero interventions during winter conditions is ever going to happen) but there are points at which it forever meaningfully changes some field of work. I think it’s there for writing code.
I would deeply implore you to not put to much faith in this technology as there are many very dangerous edge cases, including as I personally experienced, jumping red lights after coming to a full stop: https://news.ycombinator.com/item?id=48928102
> I’m probably illustrating your point but as a FSD fan
Half-and-half. I'm not denying that self driving cars (and LLMs) are improving, I'm comparing it against the standards set by the biggest proponents. But yes, I have heard basically the same thing you just wrote for the previous several major releases of FSD.
Where we agree is that, while you are a fan, you do explicitly give as an example of something you think it will never do, something very close to what Musk has promised:
"Ultimately you'll be able to summon your car anywhere … your car can get to you. I think that within two years, you'll be able to summon your car from across the country. It will meet you wherever your phone is … and it will just automatically charge itself along the entire journey."
(That said, I think Tesla's FSD will never get there, not that it's impossible. The way Musk is behaving, there's going to be a financial scheme named after him in whatever passes for a textbook in 20 years, and it won't be the positive kind of example).
What matters is the benefit it affords to drivers, not what "biggest proponents" say. Elon got overly excited in 2016, but for my day to day usage, I don't need it to self-charge on a drive across the country to get me. I need it to safely drive my family on my daily errands or weekend trips, which I now consider solved.
I'm mentally prepared for the next US administration to exact retribution on his companies, and I expect FSD will be neutered after that. Hopefully other car companies are able to catch up. I'm more of a self-driving fan than a Tesla fan, so as long as the thing works as well as what I have now, I'll be fine with it.
Yes. He was before and remains so to this day, but he did so then, too.
> I need it to safely drive my family on my daily errands or weekend trips, which I now consider solved.
This is probably unwise. As with LLMs, the statistics suggest a spikiness in the intelligence, with it being mostly good but also sometimes still making some very odd mistakes that humans would essentially never make.
As with your other comment, you know that if it gets into a crash you're responsible; while you consider this a win, I suggest waiting until the company you buy the car from (in this case Tesla) puts their money where their mouth is on quality and takes liability for crashes due to the AI upon themselves.
> Hopefully other car companies are able to catch up.
From the stats I've seen, they're much closer to the goal, relatively smoother/less spiky all-round driving intelligence. They may not be as impressive at their best, but when they fail the failure modes are themselves much safer.
The fact of the matter is that it drives reliably end to end. I don't care who's responsible or what it's called. I only care about the benefit it provides my family, which is immense. If it gets into a crash I know I'm responsible, but given that I consider it a better driver than I am now, this is an improvement over before.
I just spoke to a fried who is a headhunter and who's been trying to automate his processes for a while (he likes to fiddle and certainly has skills, but he's not an engineer). He kept trying, but it just wasn't good enough.
Now he said with GPT Work and Sol, it worked, but the key point is: all of it suddenly worked.
The problem was one of reliability, of handling edge cases. All previous attempts / model-harness-combinations were too brittle and needed too much observation and fiddling - cheaper to do it yourself.
Now he says "I don't know why I would ever hire a recruiter [the folks doing the cold outreach] again. I can focus on the candidate screening and acquiring projects, everything else is fully automated".
This doesn't come from an engineer or an AI lab, but a technically inclined power user, and I think this is where things get interesting.
You heard it from someone with no experience developing software. A lot of AI hype comes from that, even (or especially) from people in the actual business of developing software - a surprising amount of managers and adjacent or supporting roles in the field actually have very little clue about software development.
It's cool that 'regular' people can now create solutions to many small problems, and automate stuff - genuinely a step forward. Like Excel, only vastly better. But for bigger projects, real software engineers know that what LLMs do today is only a tiny part of development. And it solves it in a way that might well make the rest of the lifecycle a lot harder. It's like that saying about tools that make easy things easier and hard things impossible.
IMO it doesn’t replace software engineers, but it certainly changes the job. The problem lies with VPs and executives that rose to their positions through managing phones and printers and shit who don’t appreciate all of the work that isn’t “write code”, and those people are the majority outside of big tech in my experience.
Currently in a re-org justified by AI, AI changing the roles people will need to play. It’s disturbing how much content in the materials about the new org structure and roles and whatnot is clearly ai generated and contradictory. We’re laying off about half of 500 people.
Who are these "real" software engineers and what are they working on? Among software developers I know, the biggest users of LLMs are the seniors. I don't think any of them would agree that its usage is a tiny part of development.
Then it just becomes a new baseline (everyone have access to the same LLMs), and recruiting moves up the philosophical ladder where human can add more value. What will it be? I don't know, I'm not a recruiter.
Again I've heard this since 2022 when gpt3.5 came out.
This is like microprocessors in the 80s. Sure they double in capability every 18 months but the start is so pathetic it will be 30 years before they are good enough for everyday tasks.
CPUs in the 90s were amazing! They were over-specced for "everyday tasks." Our problem is that we overbuilt CPUs too much, so software is now written with ten unnecessary layers of abstraction because there's no real reason to simplify.
It seems to be true this time though; I have observed it myself and heard it from several experienced developers I personally know and respect. It feels like some threshold was crossed with Opus 4.5 and Gpt 5.3, where the models are now able to reliably solve certain classes of problems that were previously unreliable.
Time will tell of course, and it’s early, but inflection points do exist with progress.
Thing is. You can find an extremely similar paragraph written about Claude 4.x or some equivalent gpt. And simultaneously, many people expressing their frustration and the shortcomings of <insert any model>
“But it’s different this time” - several people, several times over the last couple of years.
This is not at all a dig at you, I’m very sorry if it reads that way. My point is these things only get truly better in anecdotes. The ways in which they fail is yet to change. Just yesterday I had gpt 5.3 generate completely awful code for the Cinema 4D Python API. Also an anecdote. But for all of the people saying they are truly intelligent and truly reason, they still make obvious mistakes, write around problems, fail entirely at architectural decisions, fail at random, generate FAR too much code.
And no amount of harnesses, methodologies, loops make much of a difference. If you listen to people on the internet they say it’s all working. You listen to people on the job and they mostly say it’s creating tech debt and a review bottleneck. Also burnout, so much burnout.
I think LLMs are mediocre. I think it’s fine they’re mediocre. You can work with low expectations. But the hype cycles are so tiresome.
I don't buy into the huge LLM hype but I certainly think late 2025 was the inflection point. Up until then, I thought LLMs universally sucked at code. GPT-3.5, GPT-4, o1, o3, Claude 3.5, Sonnet 4, the whole bunch. Each iteration got marginally better, hallucinated APIs less and so on, but my overall evaluation of them all was that they wrote crap code and were unsuitable for anything other than one-off scripts.
Then the incremental improvements did, in my experience, cross some kind of threshold in late 2025 where the things became useful. It is of course anecdotal and personal judgment. But I asked LLMs to implement a small feature in my codebase (my usual test) and finally it produced code I was happy with. They've also been able to locate and diagnose a problem based on logs. In my view it's now a markedly different level of capability than we had a year ago, though I would call the previous two years equally useless.
LLM capabilities are spiky. They're amazing at some things, and poor at others. Over time, the set of things they're amazing at has grown, while the set of things they're poor at has shrunk. If you think LLMs are just "mediocre" without any nuance, that's a sign you haven't spent the time to evaluate them in order to make an informed opinion.
> You can find an extremely similar paragraph written about Claude 4.x or some equivalent gpt
Even 3.7. I remember when it came out and people were claiming that that was now the model that was going to replace engineers. Cue Fable years later and people still claim that this one is the one.
Not shape, but even with napkin math, you can work out the conclusion. Take email vs snail mail. Online ordering vs relying on catalog. The spreadsheet, the word processor, the various CAD software.
The key thing is that it's easy to contrast the old way vs the new way and the evidence become obvious.
Okay thanks. But what I'm saying is all of these anecdotes should add up to something measurable if there's something to the idea that an inflection point was reached in November 2025, right? There's a lot of marketing and hype and motivated reasoning going on, so that's why I don't trust broadly-reported notions.
I really don't think that's how it works. Smart, experienced developers who thought coding agents were junk for most of 2025 and think they're useful now in 2026 are not saying that because they got "worn down over time".
It tends to be when the training data wanders into their area of expertise temporarily and they go “OMG, they hype is real. I was so wrong” and then a few releases later they’re on the train and furious that the skills in their domain space have not just stopped improving, but regressed. Cue someone else in a different part of the world starting the same cycle.
Meanwhile the guy who leaned in a year ago and gave up reading the output is beginning to see work grind to a halt and throwing more agents at it is increasingly not working.
You can see these tropes all over social media near constantly.
> November was, for me and many others in tech, a great surprise. Before, A.I. coding tools were often useful, but halting and clumsy. Now, the bot can run for a full hour and make whole, designed websites and apps that may be flawed, but credible. I spent an entire session of therapy talking about it.
> The real annoying thing about Opus 4.6/Codex 5.3 is that it’s impossible to publicly say “Opus 4.5 (and the models that came after it) are an order of magnitude better than coding LLMs released just months before it” without sounding like an AI hype booster clickbaiting, but it’s the counterintuitive truth to my personal frustration. [...] A year ago, I was one of those skeptics who was very suspicious of the agentic hype.
> Six months ago, in an episode of the Lex Fridman podcast, David shared how he doesn’t use AI tools to write code: he types out all his code. But things have changed a lot since then.
> In this episode, we discuss his approach to building software, how it’s changed in the last six months, and why he now takes an agent-first approach, and how he barely writes any code by hand.
> There are other questions around AI (like what the economy of it will actually look like in the end), but "is it useful" is no longer one of those questions. Anybody who doubts that clearly hasn't actually used it.
> Shock! Shock! I learned yesterday that an open problem I'd been working on for several weeks had just been solved by Claude Opus 4.6 - Anthropic's hybrid reasoning model that had been released three weeks earlier! It seems that I'll have to revise my opinions about "generative AI" one of these days. What a joy it is to learn not only that my conjecture has a nice solution but also to celebrate this dramatic advance in automatic deduction and creative problem solving.
I didn’t start using Claude Code until late 2025. Prior to that I would use ChatGPT to give me snippets of code but I was still doing most of the actual code writing. Coworkers told me in late 2025 about how they hadn’t written a line of code in “months” and just use Claude Code/agentic “whatever” so I tried out Claude Code and was pleasantly surprised. It is passable to have entire apps written by LLMs (I’ve made several that I otherwise never would have had the time to create by hand), but I wouldn’t say maintainable or easily extendable. It’s hard to be specific, but there’s something about LLM code that doesn’t look “natural”, and I’m not talking about the excessive use of comments in code. The code itself is unnatural. Functional, but unnatural. I wouldn’t want to suddenly lose LLMs and have to read through and understand and continue enhancing a codebase created by an LLM.
For me it feels a lot like generated images or video. I've made lots of things now, but those that are 100% LLM written "work" but are uncanny, weird and the details are wrong everywhere you care to look in detail.
This has been my experience has well. For code generation, you have to really constrain these LLMs on your coding style, design goals, test cases, and overall expectations. I mainly use these tools to help my understanding of the code and to generate code for very specific problems. Even after all that setup and careful review I'd say it's still a net big speed up for certain software engineering tasks.
Jason Turner gave an excellent talk at last year's CppCon explain how he thinks tools can be used to make generative AI coding assistance safer and more productive. https://www.youtube.com/watch?v=xCuRUjxT5L8
I think how I'd put it is like, it's a deterministic common denominator thing iterated on, but it's not really intuitive, it's not really "what makes sense" but rather "what would x look like", similar to how you point out it is when it comes to image generation
it's useful for scaffolding but after that I'm not sure how you could rely on it without being in the loop and directing how the code should be like
I was getting useful coding work done with GPT 3.5. I think devs saying "the models are finally good enough" this year are just trying to save face from their own previous irrational denials.
Useful, yes, sometimes, but it wasn't fully automated "Here's our JIRA board URL, fix everything that's rated 1-3 story points and in the current sprint".
Since so many people are doubting you here, here is a post from ~a year ago that's pulling the same "LLMs 6 months ago were crap, now they're awesome" shtick: https://fly.io/blog/youre-all-nuts/. There's more posts along this vein being put out from 2024 on or so.
The idea that "yeah, you said technology had improved in the past, and now you're saying it has improved again" is a gotcha just seems incoherent to me.
The context for some of these posts are yet-another-null-result coming out, saying that AI isn't having the expected objective result, with the posts defending AI by saying that it totally changed just after the cut-off the research. Evolutionary step changes are definitely a thing, but the problem isn't that AI is having such step changes, it's that you're saying that the most recent step change is the one that shows the improvement, and every study that comes out saying "no, that wasn't," the response to that is, "no, there's a new change that is totally the key inflection point!"
Or, put differently: if a study comes out next year saying they don't see major impacts from AI in 2026, will you admit your viewpoint as being wrong, or is your response going to be "no, there was a massive inflection point in September 2026 that completely invalidates the paper"?
I keep running into this as well. Someone shared a link to a study from last year telling me that AI doesn't make programmers as productive as they think. The study was obviously done months prior to it's release. So evaluating the state AI even further back.
But this didn't seem to concern them. The study said X, therefore it applies to today.
I'm just not sure it's worth it to argue with others about it at this point. Not that I'm 100% all behind AI coding, but I'm just shocked people are still this resistant.
> I'm just not sure it's worth it to argue with others about it at this point.
This is the correct response. Let closed-minded people do their thing. Makes them less competitive against you. There's nothing for you to gain by trying to help them understand what they are missing.
A year ago it was capable but required much more hand-holding. I thought it was as good as it would ever get and was still happy and productive, but only because I didn't know how much better it would get.
Nobody was saying coding agents started working in 2023 or 2024, because the category was defined by Claude Code which was first released in February 2025.
I would say that Aider is what defined coding agents. That was at least multiple months before Claude code. I remember seeing a coworker use aider for a hackathon project Adeline November-December 2024 , and it was already decent and pretty close to the DX we consider coding agents to have
Devin and Cline were the original true "agents" IIRC. Aider was the original terminal "agent" but it was't a true agent, it was sort of a hybrid agentic chat, since it had limited recursion ability (3 turns by default) and you had to configure it carefully to make it consistently take multiple turns in a row.
The timescale is well established: Late '25 was the start of agentic ai when capabilities of model + api + scaffold reached autonomous state. Any study comapring events before that timeframe is comparing apples with oranges.
Claude 4.5 was it (nov 2025?), without a doubt. It went from frequent hallucinations to highly usable with much less garbage output. If you were making demos of AI tools around this time your demo/pitch/product was saved and you probably looked like a genius.
The quality of the output is so variable. It depends on the model, “effort level”, prompting, probably even the programming language/app functionality, and libraries involved. For example, I find LLMs are best at making simple web apps. These web apps, while simple, would still take a senior engineer perhaps a week or two to create, but LLMs can spit them out inside of an hour. Conversely, LLMs struggle with things like Docker or local model stuff. Parallelization of code is a mixed bag. In these areas I think it often would have been faster for me to write the thing by hand.
I'll give a recent example: the past month I had been using Codex cloud and it was working great writing Rust modules. Suddenly last week they made a few changes and everything went to shit, couldn't get a single good result out of it. Same text box, same prompting, same model, totally different results because they changed the cloud tool's resource limits.
So whereas before my debugging time would have been spent in Rust docs looking up traits and such, these days I feel more like an AI therapist trying to figure out why it's not feeling up to task on any particular day. It can be anything from regular service outages to geopolitics that on any given day my workflow is fucked up.
In my before-AI workflow, I was never restricted from compiling Rust code because of concerns that Cargo is a national security threat. So you really have to broadly scope the notion of "reliability" with these AI tools; it's much larger than whether it can give a good output but whether it can do so consistently enough to depend on.
Here's the rub. They produce low quality work according to your rubric. If that was the universal rubric, they would already have RL'd against it, and you'd like the work they produce.
I wouldn't say it like that exactly, though you're not wrong either. People just heavily discount the value of future efficiency gains due to nuanced architectural/stylistic choices. I think this is in part because people have a limited expectation that the code will survive long enough for it to matter, or that they'll be at the company long enough to reap the rewards of the up front work.
"unless you're content with producing low quality work." - With the right guiding hand, it is a productivity multiplier without compromising quality. As a fully autonomous developer, it is a disaster.
How are junior devs becoming qualified “guiding hands” these days? If the expert with LLM assistance is multiplied, what’s a company’s incentive to pay for a junior, and how would they train to get good in these conditions?
Juniors are still useful, but not as pure "coders". The juniors you want to hire now can wear product, engineering and QA hats, and are self starters with good attention to detail and AI skills. These types of people are actually a bargain.
> With the right guiding hand, it is a productivity multiplier without compromising quality
This just reads like another variation of “it’s the user not the tool,” which is just endless runway for always blaming people and never acknowledging the limitations of LLM’s.
I’d be curious to hear how the recipients of your work enabled by the “productivity multiplier” feel about the quality.
You can't play an entire orchestra's sheet music on a single guitar either, but your playing ability still matters a lot.
I would say that as of July 2026, with the right scaffolding, you can get reasonably good output out of a LLM, or better a combination of LLMs. For example, it pays off to prepare an implementation plan with one LLM and then let another LLM check it for flaws, then again. After several iterations like this, you will have a plan better than whatever you could come up with yourself.
It often is the user and not the tool. LLMs are complicated, have nontrivial failure modes, and the user needs to steer them carefully. They might be the most complicated tools on the planet right now.
Anecdotally, the recipients of my work have become visibly more happy in the last months. LLMs are great at diagnosing subtle problems which tend to appear at Friday night only, and this is the sort of problem that bugs actual people the most.
> I would say that as of July 2026, with the right scaffolding, you can get reasonably good output out of a LLM, or better a combination of LLMs.
Totally agree, I don’t think I said or implied otherwise.
And yes can it can be the user and often even is, but when it comes to any LLM conversation I’ve been a part of it seems people think the only answer is “you’re using it wrong.” Evangelists swear it’s a 100x multiplier and anything counter to that means you’re either a Luddite who is blinded by politics or are too dumb to use the tool.
I produce code that is significantly higher quality with the assistance of coding agents, because I no longer succumb to the temptation to cut corners due to lack of time.
One example: everything I do is properly tested and documented now, even the most trivial of changes. Previously I would have weighed those tradeoffs and sometimes decided not to bother with the tests because they weren't worth the time.
> because I no longer succumb to the temptation to cut corners due to lack of time
No offense, but that says much more about the way you approach programming than about the quality of LLM outputs.
In my experience, LLMs are the ultimate corner-cutting tool. With LLMs, I now succumb to the temptation to cut corners, build something I haven't properly researched and don't fully understand, prioritize shipping quantity over quality.
Without LLMs, I have to understand the domain and the tools and ultimately my full solution (with all its warts and limitations). When I really care about the project and consider it "my baby", LLMs are out of the picture.
Software engineering is about tradeoffs. The single most important tradeoff is always time: this feature would be better if I did X, but doing X would take me Y hours, is it worth the time or should I spend that time on something else?
Not having to type in and then iterate on the code manually has a material effect on those time calculations.
I absolutely agree that you need to understand the domain and code and tools. That's what I want to spend my human time on - not typing in the code and sweating over every line of syntax.
I haven't. Around the start of 2026 is pretty widely mentioned as when they went from "this is broken slop" to "huh this is actually 90% what I would have written", which matches my experience.
The capex is still "early" for sure (i.e. data centers are still being planned and built out, we're hardware/energy constrained).
If model scaling holds out, we're "early-ish" in terms of the reliability and performance of these systems, just based on utilization of the compute from the planned capex. If we hit hard diminishing returns and we don't find architectural/data workarounds, that would put a wrinkle in things, but I suspect that the AI we have now is capable of helping us find those workarounds and keep things moving.
Anecdotally I'm seeing a lot more recruiter activity/interest now than this time last year.
But it seems more correlated with hype-cycle-stage than anything else. Right now a lot of founders seem to be convincing a lot of VCs that they can make $LOTS by replacing/changing $BIG_INDUSTRY/$BIG_PRODUCT with an agent-first blah blah replacement, and then using that money to hire more people to manage/execute/coordinate the coding agents...
Last year, by comparison, there seemed to be a mood of "software will stay the same but will require less people" while right now there's a lot of hype around "we can build different types of software or build it in different ways" and those early-stage things are in growth-mode. That guarantees nothing about how many people they'd need in the future, or their success at all, ofc.
The news that I'm getting from contacts in non-startup-land is a bit different - still layoff threats. Still pressure to use AI tools more. Mixed confidence on whether or not longer-running "agent" modes are that much more effective-without-breaking-things in legacy code if not used with care.
Is the recruiter activity you see very… Human? I’m getting a lot of emails but when I do my research, it all appears to be bots. I don’t trust any of it and I don’t respond.
However, companies have been using AI as an excuse for layoffs since well before January 2026, which corroborates the study's conclusion. (Source: https://layoffs.fyi/ai-layoffs/) There is certainly an uptick starting 2026, but that could be explained either by AI actually causing more layoffs, or by AI becoming an even better excuse for layoffs.
> This category of software may have a much more meaningful impact on work than the mostly-chat systems we were using from 2022-2025.
Mostly an impact on software development - I'm not seeing broad automation and layoffs in industries like law, finance etc. It will gradually happen but due to issues with memory, reliability and long term planning of LLMs there are real barriers. Even in software development - while it has completely transformed the field I don't think many people still believe we won't need devs in 2027 or that their amount will shrink by 50%.
I think other industries are late, but also will see less productivity gains. In software engineering agents have the benefit of very short feedback loops that provide extremely actionable insights which allows the agent to very quickly iterate towards the final solution. This is much less the case in other industries such as law and finance I think. The first pass of the agentic loop is usually something that barely works (you don't even get to see it with the latest models). If that first pass is the primary output in other fields with no tight feedback loops, then the productivity boost will be much lower than in software.
I'm optimistic that coding agents will not obsolete software developers - my own experience is that coding agents have made my work harder, because the scope of projects I can take on has increased.
My ideal version of all of this is that nobody loses their job and everyone gets to take on more ambitious projects.
I'm not quite naïve enough to assume I'm right about that though!
> my own experience is that coding agents have made my work harder, because the scope of projects I can take on has increased
Likewise. I find the types of problems I need to tackle and the challenges they represent are actually quite exhausting, too. And the agents unblock me relentlessly so I’m constantly pressed to do relatively difficult things. Either that or code review. I’m hoping it’ll only be an adjustment phase but this is the most challenging my career has been in over a decade.
This isn’t the first study showing this though. It’s pretty simple, programmers and IT were severely overhired during the pandemic, there are massive job losses now, and it’s easy to blame AI when in reality there are a lot of economic factors and AI isn’t increasing productivity as much as anyone would think.
Maybe the future will change that for very specific things, but I think people should be learning and preparing for that, which isn’t any different than what everyone has been told in every job market since the start of the Industrial Revolution.
I personally hope that AI continues not to result in a noticeable negative impact on employment and that pandemic over-hiring turns out to be the major factor for all of the layoffs.
I'm nervous that the studies which show that so far don't seem to be taking the 2026 improvements in coding and general agents into account.
What the heck would I be "stealth-marketing" here?
It's genuine concern. I do not want to live in a dystopia where AI results in mass unemployment. That would suck, even for the people who manage to stay employed.
> What the heck would I be "stealth-marketing" here?
Cynically, "thought-leaderishness".
I've been spending a lot of my time these days outlining why we can't "just make an agent for it" to CEOs who read blogs like yours. They can't distinguish a production system from a quick HTML tool from a guy whose job doesn't depend on it working.
One of the goals of my blog is to help CEOs who read it not make stupid decisions. I aim to be a counter to the breathless LinkedIn hype they are exposed to everywhere else.
No offense but wouldn’t being an AI booster run contrary to this goal? I imagine being an Ed Zitron type would be more conducive with that, albeit dishonest if you don’t genuinely believe it.
> I think this is going to have a huge impact on society. My priority is trying to direct that impact in a positive direction.
> It’s easy to fall into a cynical trap of thinking there’s nothing good here at all, and everything generative AI is either actively harmful or a waste of time. [...]
> I’m going to continue exploring and sharing genuinely positive applications of this technology. It’s not going to be un-invented, so I think our priority should be figuring out the most constructive possible ways to use it.
That's been my focus ever since: try and use whatever influence I have to nudge things in a positive direction, and help people figure out how they can use this technology in a way that is beneficial both to them and to other people.
>programmers and IT were severely overhired during the pandemic
1) I hesitate to believe that losses were disproportionately technical roles as opposed to administrative.
2) Over-hired by what metric? It's well known that hiring never fully recovered after the GFC; was the recruitment post-pandemic just bringing us to parity with where we had been 20 years earlier?
Not to say that I disagree with your following point. The AI overspending and the layoff cost-cutting are not in a direct causal relationship; both are rather symptoms of a common corporate pathology.
For whatever anecdotal evidence it's worth - that has been my experience as well. For the first time last December I noticed the harnesses performing like actual workers.
Everything impressive has happened in the last six months.
And this is why the labs cannot just "stop training and become profitable". I can't imagine they would like it if studies like this will actually become credible.
"Move fast like a blur so people can't see that you have no clothes"
Any research on the impact of AI would have been lagging indicators. It's not the researchers' fault[0], but the nature of a field moving at neck breaking speed. Remember there was a paper saying programmers were 20% slower with AI?
That early 2025 METR study was particularly interesting because participants self-evaluated themselves as 20% faster, but the measurements showed they were actually 19% slower.
All the reports of productivity since then are self-reported, or using questionable measures such as SLOC and PRs, so it’s reasonable to say that productivity improvements are still unknown.
Unfortunately, METR hasn’t been able to replicate the study because they couldn’t find enough willing participants.
Well, in addition to that you also have to consider that companies are now paying (increased) API pricing. It would be an understatement to say that my clients in the F100 range are skeptical at best regarding the gains they’ve seen compared to the costs.
This has led to many of them instilling dollar limits or demanding proof of increased productivity (not just output) with the implication being if you don’t provide value with it it’s getting taken away.
So that is to say, if they aren’t happy with the price now, how will they feel when it goes up again compared to just keeping a certain headcount?
Oh wow. You just hit something there. There are government programs for R&D where I live. Grants. Tax credits. This sort of thing. I wonder if people are trying to expense to those programs too.
>This has led to many of them instilling dollar limits or demanding proof of increased productivity (not just output)
They should have done that from the beginning - demanding proof of increased productivity - if that was their goal. otherwise they were not using their brains well enough.
And you doubly don't want to work with them, first because they confused output with productivity at first. and second, because they're parroting the productivity metric.
You only need one guess for whose pockets the productivity benefits go into.
When I was a teenager, a friend had an old gas guzzler from the 70s. I live in a rural area. One time, my car broken, he drove to pick me up to go to College.
This cost him an extra $40, in today's dollars. No, I'm not joking. That thing ate gas like a dry camel drinks water.
This is what Fabel5 feels like. Crazy expensive. 10 minutes work pulled almost $80 is usage credits yesterday. I'd be exceptionally skeptical too, on costs, if I still had the DEV I had last week, but they were also eating that kind of cash on a very-improved, but still used as a linter.
For $200+/hr, or ~$400k/year, I'd want to see a tripling of output at least. In a lot of US markets, you can hire 3 junior devs for that.
Yes, there are cheaper options. Opus, etc. But it's really over-priced, and frankly I think the real gold now is making open models fully functional. Anyone predicating their business upon tie-in with the big boys is just going to fail, hard.
This. Also I'm finding out that, after being blown away by agent mode lately, non-agent mode still kind of sucks across frontier models. Using GPT and Gemini in non-agent mode is asking for inaccurate information confidently presented as the truth. Turning on agent mode fixed a lot of that for me.
I think it's because of the lack of feedback. Humans also can't do much without feedback. E.g. I doubt most people could write 100 lines of code that works first time without even compiling it once.
Indeed. I wonder how much AI will do in industries with longer feedback loops that are much noisier. I think the field of software is a unique field in that sense.
Even more extreme than that, if you were working on a large code base where compiling it took forever, or you needed to rely on a very slow CI pipeline, it became very important to git gud and write shit that worked first time in order to deliver when promised. I'd argue if you've never had one of those moments where you sunk a bunch of time into some changes and it all compiled/worked flawlessly the first time, you're missing out.
Ah, Simon with another load of bullshit! Your own blog articles contradict your comment. But sure, keep moving the goalpost for your overlords Anthropic and OpenAI.
Also in July 2025 I was successfully using one of the earliest async cloud environments - Codex Cloud - to scrape websites.
By October 2025 the most recent Claude could one-shot a simple Datasette plugin.
What changed in November 2025 was that the level of hand-holding needed to get good results dropped to the point that it was no longer credible to deny their utility.
> Can't wait for your Summer 2027 comment where you'll claim that coding agents really only became usable in November 2026.
Oh! I think I see what happened here. I said this:
> coding agents (Claude Code, OpenAI Codex) only started working really well in late November
"Working really well" and "only became usable" are not the same thing.
Your detractors don’t care about the facts Simon, as you’re probably well aware. They’re afraid of LLMs and need to deny that they can do anything well or have any potential to take their jobs.
In their world AI never has an impact on jobs, just like AI can never be an equivalent to a human, and the goalposts move every time a goal is scored.
And yet I have felt a general code quality decrease and overall enshittification of software products since 2023 when people were already using copilot. Or I am just biased to use that to justify any overengineered piece of shit code with that because I refuse to think any sane person would come up with such contrived code and I am looking at the wrong places.
I find that productivity follows a Pareto distribution (80:20 rule) and that AI is a sharper tool in that it enables the already productive to be proportionally more productive and this effect increases as the intelligence of AI increases. So 80:20 becomes a 90:10, 95:5, 99:1 etc. I also think that corporate bureaucratic change is laggy so change there will still be rather slow. I think where change will be rapid is when the most productive employees leave the company to create a new smaller company to compete with it, so there will be a displacement of medium to large companies by much smaller ones. On one hand this greater competition of more efficient companies will result in an increased in consumer surplus, on the other hand it will result in mass economic displacement and a collapse of the tax base. I think AI has only recently been good enough to do this and it takes time to spin up competing companies so I wouldn’t expect to see this effect in any lagging indicators just yet. From personal experience, I’m well down the path of commoditizing my niche industry where I can practically give away a better version of the top tier software and still personally make a lot of money. Additionally it would be counter productive to alert my competitors to this new reality. I know I’m not the only person doing this, so this multiplied by a bunch of industries would be absolutely world changing.
> so there will be a displacement of medium to large companies by much smaller ones.
There are so many regulations EE companies follow and offer that no startup owner or worker could ever deal with. The moment your company shifts to that sort of work, you're automatically not anymore a startup (by definition).
Only if you get super smart robots that can manage a lot of bureaucracy etc. Starting from multicloud/region deployments, local regulations, accounting, taxes, laws, etc. Which is what companies like SAP (but not only them) do basically.
But if that happens we are all in a situation that we don't need companies anymore (the way we know them).
My field is lightly regulated but it is regulated and LLMs have helped massively in producing the documentation but also in reviewing the same documentation on the regulators side. This was always a disadvantage for small companies where the fixed regulatory cost couldn’t be amortized over a larger staff size and now because of LLMs it is less of a disadvantage.
I think we are living in a temporary window where the volume of bureaucracy hasn't caught up with token processing rates. In a few years it'll be increased, I'm sure.
I’m sure there will be attempts but I think our future will be dominated by separatist / spinoff societies. There is an economy of scale in functions of governance as well as a diseconomy of scale in bureaucracy. Using LLMs to make the expensive parts cheaper makes smaller self governing societies much more viable.
Why does the unemployment chart show a slow rise in unemployment for a year or more before covid started? This does not agree with the data from the BLS, which showed unemployment spiked very suddenly in March 2020. https://www.bls.gov/charts/employment-situation/civilian-une...
Also, what is “AI exposure” in 2015? LLMs hadn’t been invented yet. I did just read some of the cited paper. Essentially the quintiles boil down to use of computers, not really use of AI as we know it today. I know LLMs aren’t all AI, but the thing that’s missing is the distinction between “AI” that can play chess and LLMs that can actually do your job, which have only existed for maybe a year.
Ah, you’re right.. the dots! I still don’t understand why the smoothing makes covid symmetric when the underlying data is asymmetric. A too-large smoothing window, I guess. They must have been trying to smooth out the seasonal job unemployment rates for the labor-heavy jobs.
Hmm, I’m not seeing what you are, unless you mean ignore all data from Apr 2020 to Dec 2021. It took almost 2 full years for unemployment to return to the pre-covid norm. The spike is immediate in apr 2020, then it falls quickly for 6 months to Dec 2020, and then falls more slowly for the next year. That lump from Dec 2020 to Dec 2021 is very asymmetric, compared to Mar 2020 and the preceding year. I’d think even a year-wide smoothing window would capture that asymmetry.
(shame x is the easist way I have to post two images quickly)
dots seem to fit bls pretty well. Just to bound the effect of the spike, that is a +10% jump, so over a 3 year window it would produce a +3% bump. If they do a backwards only, it'd lag by up to 3 years. So, it's plausible/feasible, maybe not definitive.
1. Most of the productivity studies in Figure 3 about are from the 2023-2024 era. (Which is why as some comments note, Copilot is actually way up there in the numbers. Note that this was from the era of spicy autocomplete and long before coding agents exploded on the scene.)
2. AI adoption at work is actually very low: even though 50%+ of Americans currently self-report (major caveat) using AI at least weekly, they use it for only 6% of work hours. (You can play with the charts here to see this [0]) This is what the recent Google study [1] called "broad but shallow use."
Notably, the same surveys find time savings of 2% of working hours, so a whopping 30%+ productivity boost per hour of AI used. And this is across industries. Despite being self-reported, it does line up with many of the other controlled studies (see TFA and [2, 3]). Some economists suggest that even with this low level of adoption, we may already be seeing the impact on labor productivity at national-level aggregate statistics! [2, 3]
As I said in another thread, my concern is that the impact on jobs is only beginning because 6% of work hours is a very low number. However given how useful people are finding it based on self-reported, micro- and macro-level numbers, adoption is only going to up, both in breadth (more people) and depth (more tasks). I fear the impact will happen gradually, as adoption inches up... and then suddenly.
Recently poked around the job market to see what I qualify for in this day and age. Working as a solo builder in my org I would say that I have done enough in the last 18 months to consider myself “with it”.
What I found was pretty brutal. Companies asking for 4 years of agentic AI experience… pardon?
Then it hit me.
Oh they are all making shit up now and have no bar that anyone can hit because they are believing in the hype without understanding the fundamentals.
GREAT. Even as I climb the AI-Native ranks, I apparently am unqualified for any AI-Native job.
If decision-makers put up a front of recruiters and talent acquisition "specialists", don't they send a very explicit message that going past those is unwelcome and won't be considered?
In practice, skipping the line is THE approach. How many times do you hear people say to fill out online job postings? Never. It's just scaring off good/bad candidates and bums who dont have an employee referral.
It's not even cynical to say a lot of postings are performatory compliance steps when they've already got internal candidates lined up.
If you were screwing around with Auto-GPT in 2023, you'll hit four years of "agentic" experience next year.
Of course, there's not much overlap between that and the way it works now. But then again, I have the same feeling about last year and this year... (e.g. Anthropic just deleted almost their entire system prompt because the models have common sense now.)
That being said, I think there's still value, even today, in playing around with the older models from time the time. (Or with very small recent models, which have similar limitations.)
Some of the habits that teaches you — i.e. careful context management and well crafted examples — do translate well to the modern environment, and give you performance gains and cost savings even with newer models. (In a word, whenever possible, show, don't tell.)
I also lament the loss of the base/text models, which were extraordinarily interesting and fun to play with. But that's a separate discussion :)
There’s a big problem with hype now, that’s for sure, but what you’re describing also reminded me of the companies that were asking for 5 years of Java experience in 1999.
I guess maybe they were all trying to hire Gosling and his colleagues? …
Just in case, because a lot of people don't know this... When a company lists job requirements they aren't really requirements. Don't skip a job because it says you need X years of Y but you only have X-1.
They're basically writing down a wish list. They don't expect to get it or necessarily even care that much about some of the points.
Also "X years of experience" isn't really asking for literal years. It's a proxy for skill. They mean "as good as the average person who has been doing this for X years". If you're really good at it and can demonstrate it, that's good enough.
yeah the lead lecturer on my data engineering masters once told us about a job posting he saw asking for 8 years of Hadoop experience, even though Hadoop had only been publicly available for 5 years at the time.
it's often HR / hiring managers throwing some numbers into a text document based on what they've heard is important for the role, not what you'll actually need for the job.
similarly, something like "has previous experience with kubernetes" doesn't usually translate to "knows absolutely everything there is to know about kubernetes". it means "you've used it at least once, ideally more, but can at least talk about when / how you used it / what problems it solved and could probably get up to speed on it fairly quickly when you join and/or in the time before you join" (kinda writing this last bit about an interesting job posting i saw that i've been talking myself down on and i really ought to be doing the opposite).
If a tech company can't even do basic math — or rather, openly cares so little about accuracy — does that sound like a place you would want to work though?
this sort of thing happens more often at larger non-tech enterprises, or places that have a type of "software team" that is not the main focus of the org.
> Companies asking for 4 years of agentic AI experience… pardon?
Not that I am trying to excuse it, but this is not a new thing, nor specific to AI.
Job listings that ask for X years of experience where X years is sometimes literally longer than the technology has even existed has been a staple complaint of developers over my entire career, and I'm old af.
A 10x engineer only needs about five months of experience.
So, leave college/uni with your "Desmond" (1) in comparative pornography in Feb 2026, buy a PC/Apple and by now you will be writing Windows Entra 2027 on your own.
bro, tech hr were asking 6 to 8 years rails experience before dhh (rails creator) was even born.
This has been almost a meme on hacker news for some time.
You can google it via hn dot algolia dot com by using the right keywords.
Of course, i exaggerated it a bit, just like a lot of startups and vcs pimp their stuff, just that they do it much more, and they do it for money, while my mine was for fun. ha ha ha.
The biggest impact of AI at my job are LLM code generators for software engineers. That's only scratching the surface of what AI can do for enterprises. My company doesn't yet have the orientation, inclination or technical expertise to unlock the true power of AI for our business goals. We're stuck buying packaged solutions from third parties and aren't integrating layers of intelligence in our environment.
I doubt we're unique. Chat bots are useful. But it will take years, possibly decades for work to transform to due to AI. Probably longer for everyday life. The diffusion of new technology, even something as profound as AI, has to fight the friction and realities of the real world. Always has.
Big caveat to the productivity claim is that it’s concentrated in lesser experienced engineers and vanishes or goes negative with highly experienced engineers.
This intuitively makes sense and generally agrees with my experience. LLMs move your baseline towards the mean. If you are below average it improves and if you are above it hinders.
But also in my experience, my memory retention of the work done with an LLM is worse than doing it myself self. This leads me to believe that the lesser experienced engineers are not gaining experience!
My tinfoil hat says that this is what tech CEOs want. They want workers that are low skilled and can be paid less.
>My tinfoil hat says that this is what tech CEOs want. They want workers that are low skilled and can be paid less.
My company is small, but my opinion at the moment is the opposite. Low skill workers don’t notice the flaws of what LLMs produce, I am no longer interested in hiring any juniors, as all they do is copy and paste LLM output without thinking, which anyone could do. It is the senior engineers who are most valuable to me these days.
And as a senior engineer, I'm not interested in cleaning up vibe code. When I review a PR for a junior dev they learn from my suggestions (or I tell their boss they're not learning). Having to correct an AI is a waste of my time.
I hope you do know that your comments are straight up being pasted to an LLM that addresses or counters those, not the junior dev. They are likely learning nothing other than how to increase their PR counts for managers to see.
That's not how you should use it and LLMs aren't that bad anymore. Either you just accept the quality level of LLM code wholesale, which means accept mediocrity in exchange for volume, or you at least leverage the LLM to improve the process surrounding the production code.
> My tinfoil hat says that this is what tech CEOs want. They want workers that are low skilled and can be paid less.
I think it's useful to divide companies doing software development, as well as specific jobs within a company, into two groups:
1) Companies where software is the product - where what you are working on is directly what is making the company money, and differentiating what they sell from what other companies are selling. These are the jobs you want, and where you will be valued.
2) Companies where software is only a component of the product, regarded as non-critical, or not product related at all - just internal IT systems. These are NOT the jobs you want!
In a type 1) job the company (unless it is run by idiots) recognizes that better developers = better product = more profits. They will seek out experienced skilled developers and pay them what the market demands.
In a type 2) job you are not regarded as a profit-generating asset, but rather as overhead - an expense to be minimized. The company will be looking for the cheapest, least experienced people they think can do the job. Maybe they will outsource, and/or nowadays try to use LLMs as a way to avoid needing to hire better quality developers. The development work may still in fact be demanding and require skilled developers to be done well, but if the company mindset is that developers are an expense not an asset, then they will try to get the job done with the cheapest labor regardless.
So, yeah, if at all possible don't work on things like IT systems or on products where it is not blindingly obvious to management that software quality is directly related to profitability.
this is useful and thx for that, but there are more layers and variations. In some companies, marketing CxO has the budgets and therefore the power, while minions execute the vision. In other companies, there are capital assets, physical assets and historical non-public relationships.. In those companies, the security of control, and the power of those that do the control, guide the boardroom. Software is involved in basically all of these, but common goods software from a vendor plays a large role. Just these examples from imagination lead to many permutations of "how software devs are treated in the company"
> my memory retention of the work done with an LLM is worse than doing it myself self
Do you think this is more because you are less close to the work? Or because there is simply more work being done/more to remember, so you're remembering a similar amount, but a smaller fraction?
I believe it’s mostly due to closeness to the work and actually turning the problem over in my mind. When I use an LLM to code much of the context is lost and my understanding of the solution is diminished. The longer I use it in a code base the less and less aware of the code base I become.
This is also born out in the research which has demonstrated reduced retention of content when authoring was facilitated by an LLM. Over the long term, I believe this is widen a chasm between experienced and inexperienced engineers.
While not all CEOs will feel this way, it will not surprise me that knowledge workers will be treated as disposable, even if their work is invaluable. If they can hire a junior escort for the AI they will. That junior will also be scale goated when things go wrong.
People gloss over memory retention way too much. You become like a manager who knows about the overview, but not the ins and outs (to a lesser degree).
Maybe it's only me and my inexperience, but only reviewing doesn't make me as "in" as doing it myself.
> My tinfoil hat says that this is what tech CEOs want. They want workers that are low skilled and can be paid less.
Guess it worked then. I'm happy to take a paycut to never deal with vibecode and toxic environment where every estimate is borked due to agentic coding.
Where are you getting that from? It's a productivity tool. Someone who knows what they're doing can leverage LLMs to improve their productivity. Someone who doesn't know what they're doing will be able to increase their output but not do it as effectively and with lower quality.
If anything, it now requires more skill to maximize. If you have strong product and engineering experience you can get even more out of an LLM than one of those skills alone.
I'm sure Anthropic would love for everyone to believe that soon they will be the sole provider of "skills" for the entire planet and every single business transaction is done as a part of the subscription they offer - people sell AI generated content and other people's agents buy that - they're the entire world economy. I'm sure such narrative would help them reach infinity market valuation for the upcoming IPO.
Organizational inertia is a real thing. There are still fortune 500 companies with internal bans on AI. A lot of the answer to "how much impact has AI had" comes down to "how much have we even attempted?"
In my workplace, we're going to decline to renew some software subscriptions because a non-programmer vibe-coded their replacement in a week.
The impacts are here, they're just not evenly distributed yet.
> In my workplace, we're going to decline to renew some software subscriptions because a non-programmer vibe-coded their replacement in a week.
The impacts are here, they're just not evenly distributed yet.
Interesting to see the impact in the long term when battle tested software gets replaced with vibecoded variants by non-programmers. Does it increase data breaches or quality actually goes up?
I think Sturgeon's law would tell us that everything will stay about the same.
But in reality, a lot of corporate software exists just because there are plenty of companies who are afraid of owning code. They don't want to maintain any in-house coding skills, and therefore are willing to buy literally any vaguely-relevant CRUD app that the manager heard about at the conference. I don't think replacing that class of software with vibe coded alternatives will be any worse, because the bar is starting on the floor.
There are entire software categories that consist entirely of code that is only one or two evolutionary steps away from some engineer's spreadsheet originally written in 1995. One fine example I work with has changed its backend database 3 times in the past 4 years. Their most recent decision to use mongodb came with the questionable decision to store json as a raw string literals complete with bizarre escaping inside a database literally designed to store json-shaped-objects.
I don't think Opus could store data that poorly, even if the end user prompting it didn't know what they were doing.
One of the upside I can foresee is the return of single payment software. No one is going to be paying subscriptions when they can LLM themselves, but paying someone else to LLM for them would let them shrug off accountability for malfunctions
The other day, i was fed up with my mac webcam software. I just needed to fix being washed out. Every piece of software I found cost money, several were a monthly subscription.
Over lunch, Claude made a minimal toolbar app that lets me adjust down the brightness.
Often, you don't need battle tested. And you don't need a bunch of features.
The ai could have shipped my video feed somewhere I suppose, if I were unable to read its code.
Becomes a problem when the next version of Mac OS changes the api, and you have to dig up the code, install dependencies, recompile and fix it. Meanwhile potentially breaking something else.
To me, this is inline w the book BULLSHIT JOBS, and that class of job that was really like a few hours a week but 40 hrs pay. That type of job should be automated and the person removed. The exception is going to the person who complained to management that it’s a 2 hr a week job…. Give me more work.
> In my workplace, we're going to decline to renew some software subscriptions because a non-programmer vibe-coded their replacement in a week.
This is curious to me, as at my workplace we never had such subscriptions for small- to mid-size stuff and always built the corresponding tooling in-house.
> There are still fortune 500 companies with internal bans on AI.
And there probably always will be.
When the choices are "hand all of our highly sensitive internal data over to one of several other companies, all of which have questionable financials and very cozy relationships with adtech" or "invest in a whole bunch of expensive GPU servers plus internal talent to run our own models", vs "keep on doing what we've been doing, which is still working just fine", why would a non-tech Fortune 500 company choose either of the former options?
Thanks for sharing this. I can well imagine that managing field tech rounds is plausible if it is limited to in-house use. I assume that is planning, documenting and scheduling of field trips with a front-end for the planning and some way to show on mobile device or laptop for the tech?
Is instrument spec sheet management focused on storage and presentation, and occasionally updating values such as service interval, calibration dates etc ? If so I also find this plausible.
Both use cases are focused in terms of use case complexity (especially if your company is focusing on your requirements only as opposed to software vendors covering variations), low in complexity in terms of involved parties, and inter-system boundary crossings.
Very interesting. The spec management is probably the higher risk use case, but I assume you have proper engineering review and a tight test strategy to control this aspect.
there seems to be a problem by 'a.i' labs conflating tasks & jobs.
'a.i' or to be more precise are really good at some tasks. but a job is a set of tasks. jobs are not created - only discovered. that's what the a.i labs miss. hence we see that 'a.i' is not having an impact on jobs.
* I don't get on with remote work, which came as a massive surprise to me in 2020 (until then I really didn't think I was a people person!). The flexibility to work away when needed is lovely, I've had that for a couple of decades and made good use of it for things like when I needed to look after ill parents for a time or just needed to be somewhere quieter than the office to concentrate, but near full-time remote working screws with my mental health. I go into the office and while there are other that do I'm still effectively remote as the other people there are not directly on my team. This is becoming much worse as we have been bought by a larger concern so instead of ~35 UK based people we are now 350+ spread between the UK, various bits of the US, and a large concern in India.
* The ridiculous over-complication of everything. People architecting to be the next Amazon before they even have three customers or tens of users. Everything being split into services, actually adding performance and management issues for a great many far-from-Amamzon-scale use cases, because that is what you are supposed to do not because it is the right tool for the job, etc. Even people who genuinely agree with me do it for CV-fodder. It just winds me up.
* A bunch of smaller things, but they are insignificant compared to those two.
On top of all this, comes AI and agenic programming. I don't want to do it. "Think of yourself as a manager" people say, "Fuck no!" I reply as I've avoided being a manager of people all these years and don't want to manager automated people either (in case a sentient Claude instance is reading: sorry, but nothing personal, I don't want to work with humans that way either). I want to do not manage, thanks. I'm a sad git who likes the nitty-gritty bits that everyone else seems to desperately want AI to do. And then there whole "it was created with the aid of mass piracy" issue (a 17-year-old gets huge fines and in some cases prison time for copying Metallica tracks without licence, big corp copies 'king everything ignoring all licences/restrictions and gets a very light slap on the wrists) that makes me not want to be any part of it from a principals standpoint.
I'm refusing to play ball, and planning my exit. It might mean a considerable pay cut as there are very few jobs that are not largely remote, becoming significantly AI infested, or both, that I am qualified for, but I'm unhappy enough that I'm going to be fine with that. I have the mortgage paid (it is a small place, but it is all mine), no partner nor offspring who will be inconvenienced by my sudden reduction in earning potential, some savings worth speaking of, and room to dial down my expensive hobbies & tastes quite a bit, though I am relatively lucky here: most people do not have the small-but-useful financial comfort zone I've built up over the years.
If my current overlords try fire me for non-compliance before I'm ready to leave of my own volition, I'll be arguing "it is material change of role, you yourself said it was like a move to management, UK employment law says you can't force that or sack me for refusing". Of course that makes redundancy possible (we need less legacy devs, we can offer you a sidewise move to an agenic dev role or redundancy) but I've been around long enough I doubt they'll go there due to the expense (if they do, then thank-you-very-much!).
There are a few tech options I can explore that I might be qualified for, if those don't pan out then I look forward to the last decade-to-decade-ana-half of my working life in hospitality, or maybe as a hospital porter, or whatever. If I leave soon, I might latch onto a good path elsewhere before the mass displacement of dev & other tech workers is clamouring for the same things then I'll at least be a step ahead of you all there!
AI already can do basic stuff pretty well. Think about people doing robotic work, ugly internal tools in big corporations and other. Decent engineer managing and reviewing AI output can a lot without big cognitive load and bunch of people which were needed in the past.
Employees have no incentive to meaningfully implement AI to increase productivity, and if they do so, they have no reason to share it.
If AI means job cuts and not using AI means job cuts but no one really tracks AI impact that means performative adoption of AI is safest, and real gains are to be sandbagged as innate magic hand waving.
At my work I'm one of the few who says "I made this with Claude" and the near impossibility of using AI with internal email etc for security reasons means AI use is one-shot wonder oriented (for me, in my experience).
If AI usage was incentived with actual bonuses and praise it might go over better. So far I haven't seen that. Its just implicit threats.
This was observed on first week for me. In the employee room, most people were all crazy about LLMs, looking at the prompt results in awe, "I can do my day in 1h easy..." . By the end of the week, group meetings were much more tame "yeah I can go a bit deeper than I would otherwise, surely saves me a few hours a week". And after the meeting was over these employees were already saying that they would never report the real productivity boost, because that would be suicide.. more trash talk will fill the void instead.
I don't think we should fear the big corps shrinking over night.
What we should fear is the startups.
Legal startup demonstrates they can do legal work at a fraction of the price. New generation of banks can underwrite at a fraction of the cost, or new insurance companies can insure much cheaper.
This will force the incumbents to reduce head count and use Ai.
Yes, I’ve found my coworkers won’t admit they had AI write their code in any sort of group setting yet they’ve all admitted it in private. The implication seems to be if AI is writing your code you’re lazy/incompetent/bad/not working anywhere near 40 hours/week. I for one would hate for our bosses to figure out we do probably ~16 hours of actual work per week. But we all “get our stuff done” (thanks to AI).
AI is only replacing low-level work. It will eventually move on to high-level work.
One missing point: careers that require communication, especially person-to-person, won't be replaced by AI. Why? Because people don't like to speak with AI when they have a hard pain point to solve. It's not about intelligence; it's about trust and relationships that AI cannot replace.
> careers that require communication, especially person-to-person, won't be replaced by AI.
A crucial point is that the amount of these careers might reduce dues to structural changes of the job market.
Just: yes, SaaS companies can do their work faster and can downsize. But a sizable portion of SaaS companies get to downsize entirely as their entire purpose is disrupted.
One thing on the first graph that jumps out, is that ALL unemployment has gone up/down in same ratio's. Maybe the current job market is not about AI, but all the other variables in the economy that has always driven unemployment.
Maybe the scary thing, the economy itself is suffering which is causing the unemployment, not AI.
this is not happening because ai is so good it can replace workers. its because its good enough to look like it can replace workers to a non technical manager, and that gives execs an excuse to fire people (or not hire them) so they can show better margins and get a bonus. even if they know it kills the company long term.
the root of the problem is the same as most other economic problems in this world. financial capitalism is designed to reward short term profits, and shareholders create an asymmetric incentive where ceos dont really get fired for doing too many layoffs but can be fired easily for not doing enough.
I am just having a hard time with all this AI stuff. The other day I got some feedback on one of the products I own. It was a small tweak to an anchor tag to add a title and change the color for just that element. I made the change in the web inspector, showed it to the user, they said yeah that’s good. I walked over to one of my engineers and asked “can you make this change and get it in dev, I just pinged it to you”. No shit they turned around and pasted what I gave them into copilot with claude and asked the agent to do it. Well it actually didn’t do it exactly the way the customer wanted, so it actually took 2 tries with claude. Incredible.
Prior to the weaving loom and sewing machines, a factory might be able to make like 20 t-shirts a day with 50 workers. After mass production, factories generally employ the same amount of workers, but people aren't hand-sewing things. Instead roughly 50 people are producing thousands of tshirts a day leveraging giant machines.
Most of HN recognized the "We're firing people because AI makes people efficient" as one of the stupidest sales pitches ever and the CEOs that fell for it are just poorly ran companies that outed themselves.
AI is just another cycle in technology that is genuinely useful. The companies that are going to jump the gap are those that are hiring to use this new skill. If 10 workers pre-AI yields you 10x, and post AI yields you 100x, you don't cut down to 1 worker so you can keep delivering 10x. You invest, ruthlessly train, hire, and surge forward and leave your competition in the dust.
If there was 100x on the table for you, why weren't you already at 100x scale just with more headcount? That is the other side of this coin. Demand is a factor. Productivity goes up? Great, if there is demand to satiate that you can meet. I doubt that is the case otherwise you'd already have scaled to meet it if it were actually on the table for the taking if only you could output more volume.
"factories generally employ the same amount of workers"
Not really. The stats should show that factory workers as a percent of the workforce has been declining.
Thw leadership at my company has said they plan to do more with the same people rather than lay off. But they also seem to have reduced hiring and are removing certain types of roles.
Agreed. When tools evolve significantly the old production process has to be reworked into a new process and that takes time. In complex production lines that will include changes to subcontracting, supply chains, and even finance models. The changes arising in software production today will take years to stabilize into a new normal, even longer if the tools continue to advance in capability as LLMs are doing. And many software-dependent companies will fail in that time for moving either too fast, too slow, or most likely, too short-sightedly.
The sewing machine analogy is a good one. The problem I'm seeing is leadership who sees the t-shirt company succeed with sewing machines, so goes out, buys 100 sewing machines, then puts them in the kitchen because they are a bakery and gets confused why the bread production isn't improving.
Because AI can do some things. Not everything. Applying it to the wrong problem reduces productivity.
Anecdotal: I used to beg my boss to get some help. I wanted 5 engineers and would have been upset if anyone had been hired that wasn't an engineer until I at least got 1-2. That was up until about 3 months ago.
Now, I don't want any additional engineers. Not only because I don't need them anymore, but because the prospect of having junior engineers using AI is absolutely terrifying to me. I can't eyeball their code to get a sense of how good of an engineer they are anymore, and I can't possibly review all of their code because of how much code AI can output now. So they're going to be outputting a ton of mostly high-quality code that could be making horrible mistakes that are much harder for me to catch now.
I don't feel like reading an article that will probably be out of date in 6 months, but from what I've seen, if agents keep improving at this rate, 80% of SWEs are going to be looking for new careers in 5 years.
The problem is not necessarily job losses as such. When artisans were replaced by factory work, it took a long time for automation to actually reduce the number of workers. What happened quickly was that the asset of the artisans -their skill and experience - was replaced by an asset owned by the factory owner
-the capital equipment. That converted workers from having guaranteed work, and a degree of autonomy, to being a replaceable commodity dominated by the factory system.
This is the real threat today. The growth of the middle class was essentially the growth of jobs in which the workers have human capital in their skill and experience again. That experience is being slurped up by LLMs and other models.
The question is whether "LLM operator" is really going to be a profession which requires scarce skills and experience, or whether many companies will be operable with less well-rewarded workers. I think that those who think there's an obvious correct prediction here are overconfident.
Probably both. We were already hitting limits in terms of finding software that could generate returns; without public funding or enormous regulation the possibility of finding new jobs with a degree is lower than it's been in a while.
Seems like all companies think that their products are good enough and can’t offer more, as they always want to remove people instead of offering better and bigger products.
I wouldn’t build my product with an LLM owned by a company at the core of it. That feels way too risky, it’s almost like building on a platform you don’t own. I expect we will see way more entrepreneurial growth once local models become good enough and the hardware to run them becomes more affordable than it is now.
(Though I’m biased, I personally feel it’s an anti-human technology, the world would be better without it)
The problem is that these two statements each have massive implications, so instead of treating these findings as point in time snapshots they are the whole ballgame and should be explored in depth.
> I find that productivity follows a Pareto distribution (80:20 rule) and that AI is a sharper tool in that it enables the already productive to be proportionally more productive and this effect increases as the intelligence of AI increases. So 80:20 becomes a 90:10, 95:5, 99:1 etc.
> Big caveat to the productivity claim is that it’s concentrated in lesser experienced engineers and vanishes or goes negative with highly experienced engineers.... LLMs move your baseline towards the mean. If you are below average it improves and if you are above it hinders.
reply