Is this sustainable?
jamiehurst.co.uk
jamiehurst.co.uk
This kind of senior engineering role really depends on the type/size of an org. I've had jobs like this on & off and generally don't stick around for long. There have always been high-impact hands-on-keyboard senior roles that involve coding most/every day..
> The other thing that gave way was thinking time. There's very little of it in my working day now. The productivity gains from AI got captured by output volume rather than output quality.
I actually see this externally from b2b vendors I am a client of. Companies that used to churn out X new products/month are now pushing 4X products but they all suck. The quantity over quality market is going to produce new opportunities for others.
> Three years ago [...], the process was familiar: write a proposal, get feedback, iterate, build a small PoC to demonstrate value, get a team assigned to take it to MVP, ship something fully featured and integrated with the rest of the platform six to twelve months later.
This to me smells of large, slow, very political organisation where actual work gets done at glacial pace. The increase in speed is probably not due to LLMs, rather to the fact that this person now has an excuse to present working products while before, by their own admission, they were mostly dedicated to producing corporate slop.
once the linkedin anti-ai hype train starts in earnest that’ll be when there’ll be money to be made.
monkey see, monkey do.
The rest will be out touching grass.
at that point there’ll be some money to be made for the supposedly “replaceable” software engineers they’ve been shitting on this whole time.
i’m hoping my bet pays off, cos i would like nothing more than to rinse these people for everything they’ve got. it’ll be payback time baby.
> The cost of building has collapsed, but the cost of aligning organisationally has not. If anything, it's gone up. When three different teams can each produce a working solution to the same problem in the time it used to take to write a proposal, the bottleneck moves from engineering to coordination.
We're still figuring out how to productively use coding agents as individuals, the next challenge is figuring out how to productively use them within teams. Coding agents reduce one bottleneck - producing working code - but that just moves the bottlenecks elsewhere.
(Note I said "working" code and not "good" code, that's a whole other thing.)
I thought labs would have pushed collaborative steering by now, but I guess people got so TUI pilled they haven't even considered the possibility.
I've taken to linking to mine in commit messages, e.g. this one: https://github.com/simonw/simonwillisonblog/commit/e781e4eef...
(My favourite feature of the new Codex desktop app is that it has a comprehensive "Copy as Markdown" feature, which I can then paste into a Gist.)
I'm sure the labs are working on this sort of thing - Managed Agents are probably the closest?
It’s a distributed agentic system on Temporal where all inputs are signals to the workflow. And then each agent has its own GNOME desktop either in a K8S pod or KubeVirt VM.
The biggest problem was context and ownership. The more we kept steering the model the worse it got until eventually it couldn’t complete its task. And on the ownership side, no one clearly owned the outcomes so it was just there… generating slop.
Not sure what adding the ramblings with an LLM will add.
One simple observation is that using agentic AI coding tools is not a multiplayer game currently. It's all one on one. I think this is a problem because a lot of development work actually involves multiple people. And to be effective as a team you need a shared context. If your agent has a different context than mine, we're not working together effectively. Having meetings to work around this is not a great solution. I call them synchronization bottle necks. Because that's what a meeting is. We drop all work to gather in a meeting where we share information and take decisions. And then once we are synced up, we continue work. OSS development happens mostly asynchronous tools. There are no meetings typically and hence less bottlenecks.
But in large organizations, all decision making bogs down on communication overhead. Miscommunication and misaligned people become problematic. The bigger the team the worse it gets. It's an exponential. You end up spending most of your time in meetings where you are collectively blocking each other from getting stuff done. It's why startups are so effective when they are small and agile. With AI we can move at that speed as individuals. But not if we are spending most of our time in meetings trying to brief everyone on all the amazing stuff that happened when you were talking 1 on 1 to your agent. It's a lot of context to dump in a meeting; it obviously does not scale.
The path to a solution might be very simple: why not have agents show up in team chat tools: make this properly multiplayer. Or, better, make the agentic coding tools proper team chat tools. Now we have shared context.
And then maybe rethink the whole process while we're at it. The whole business of having standup meetings is a good example. Or planning meetings. All of that is valuable context for LLMs. So, make sure they are there or at least have access to the transcripts. Use AIs to summarize, provide digests, track issues, update statuses, etc. They are really good at the detail work.
And you could do more radical things: instead of a scrum master, why not have a scrum agent? Why create issues manually? There's a lot of repetitive around issue and backlog management that POs still do manually.
We haven't really scratched the surface of how AIs are going to change development. Some of this probably won't work. Or it will work completely differently. The point is we have yet to try most of this stuff and figure out better ways of working.
These are words that humans use, but that Claude loves to use in a particular way, the kind of way used in this article. It particularly likes the phrase "The honest version".
Add "X is real".
Why? You are already on hn which filters content before you read it
I do not mind at all that someone creates an AI Image to decorate a blog article.
I'm also not directly offended if someone is using an LLM to write text.
And i find it a valid argument, that people who post on HN and get enough votes for content, people read and liked, that this can be enough as a filter or quality gate.
In another comment they mention that there is some kind of mechanism of sometimes showing new in the overview. But besides that, its filtered
I am rewatching Stargate SG-1 with my children right now and cannot read this comment except in the voice of Teal’c.
In the previous episode the team was in 1969 on a hippy bus.
My advice is to chill out a little…
I was listening to some obscure band you wouldn't know with a few chicks I met roller blading last week, and I can't help but read your comment while imagining the first few notes of the second song playing. After that we watched The Simpsons and my advice is to not eat a cow, man.
The truth is, I still don't know how to put it more nicely, but please have the generosity to take my lampooning of your comment as the time honored tradition of whatever you call it when you point out the flaw in something by driving it to an absurd extreme. It does have a place, but my eye-rolling "being gobsmacked" was a bit too much indignation. tbh when I wrote that I thought I ought to get flak for that, but fuck it, I'm annoyed, I'll say it anyway. But I didn't get flak so now I feel the need to calibrate it myself a little.
In short, I think you're wrong, that comment was bleh, but so was mine. Have a great weekend and much fun showing your kid stuff you both enjoy.
> The cost of building has collapsed, but the cost of aligning organisationally has not. If anything, it's gone up.
…Oh.
> What gave way is the human-focused work. Mentoring is the clearest example. I have less time for 1-2-1s than I did three years ago, and that isn't an accident, it's a choice I've made under pressure.
Ohh.
> The other thing that gave way was thinking time. There's very little of it in my working day now.
Ohhh.
Followed by....
> Senior engineers in AI-forward orgs are doing more leveraged, more hands-on, more meeting-heavy work simultaneously, with the human-focused parts of the role paying for it. The build cost collapsed, the alignment cost rose, the thinking time disappeared, and the productivity gains got captured by output volume rather than output quality. I
What is the job of a senior/lead engineer if not to take the uninformed hype chasing of the senior business, and deploy it in a way that makes things better?
I can't help but feel this senior engineer is talking far too casually about how - under their watch as senior AI engineer chap - engineers spend more time on throwaway code, have less personal development/1-to-1s, and didn't improve code quality. They haven't even mentioned the added financial token cost.
The only thing standing in the way of greedy hype chasing CEOs and a post apocolyptic wasteland is engineers taking their crazy requests and not making the world worse, and it sounds like the author has failed here. I think it's very positive and frank to share their experience, but I'm surprised they don't seem to see their role in it.
Nothing.
Either you quit outright, or ride the flames to the bottom, depending on personal ethics to cash-out potential ratio.
It's not an engineer's job to fix upper management delusions, and engineering is poorly equipped for that in any case.
What is the job of engineering leadership in your mind then, if not to take requirements from the business and convert them into technical solutions which improve things for the business? The requirement here was "AI to improve DX" the outcome was developers lives are worse in many creative and future impacting ways.
I sincerely would love to know if the author or their employer consider what they've done as successful or not.
Do you think the author of the original article should have quit, rather than tried?
On the other front, people are saying that NVidia can't deliver stable drivers for like 15 months and they don't want to take software updates at all, they are more happy with last year's drivers.
I think this is a black swan event in the industry. A lot of people already suffered and more people will suffer still. Industry is going to change for sure, but probably not in a way that you would expect. Black swan simply doesn't work that way, it doesn't change industry in a good way, hence black swan.
[1]: https://www.researchgate.net/publication/402969772_The_AI_La...
Thus it can change industry in a good way or a bad way, because the black swan is unprecedented and unpredicted, its consequences and their nature is unsettled.
The only thing that is unsettled here still is how many more people will lose their jobs and how much cumulative loss prisoner's dilemma will generate.
I saw random people on Internet suggesting to piggyback this disaster and dip into the crazy money that it is "generating", but in a zero-sum game somebody has to lose.
All that said however, probably the greatest black swan in terms of unpredictability in human history was that there were two freaking big continents in between Europe and Asia when you try to circumnavigate the globe.
on edit: probably should be "the greater the unpredictability and the more things the unpredictable event connects to the greater the negativeness seems to be", because obviously the platypus while completely unexpected affected basically nothing at all.
Technically black swan might not be negative, but for all practical reasons, "black swan event" is what people call impactful unpredictable negative events and they expect those events to be negative. In other words, it is a synonym for "a disaster" of sorts, only "economical".
This is how an economist would call a disaster.
https://theconversation.com/friday-essay-a-rare-bird-how-eur...
An observation consistent with past observations would be a "white swan".
But maybe that’s not the intended meaning either? It’s an interesting expression.
When I left my corporate engineering job wayyyyy back in March, there were engineers and engineering leaders going off and getting a lot done, individually or in small teams. But project management and QA couldn't keep up with it. Managers resorted to turning their tokens loose on Jira just to try to make sense of it all (which, ironically made them the first to hit their token goals on the dashboard every week, and brought Jira to it's knees).
And, even worse, the junior engineers had no idea what was going on or how to get involved in anything.
The result was an increasingly chaotic mess.
Coordination has in my experience always been the big bottleneck in getting anything done, it’s just not hurt so much because everyone expected a feature that could have been done in a fortnight to take months.
> Coordination has in my experience always been the big bottleneck in getting anything done
I work at a large enterprise you've heard of. They're currently re-organizing the product area to remove currently-static two pizza teams into an amorphous blob of feature-oriented teams. Once the feature is complete, the team is dissolved and the engineers re-enter the pool, tasked with new features.
All that to say, I think you're right on the money with your assessment.
I am not sure if that's a good thing for Claude, or an indictment of Jira.
If only this was limited to engineers.
I'm seeing this applied to every role across organizations. The designer that gets heard is the one who can vibe code the best. Same with strategists, writers (seriously), and other people for whom "coding" is not remotely their expertise or function. This exponentially accelerates the "cost of aligning organisationally" when you have not only several eng groups but every single person in your team regardless of role proposing their own solutions, that as an engineer it is now your responsibility to babysit (or as we are terming it, "harden").
Meanwhile, the quality of creative work suffers because instead of working in a surface that's designed for their trade and process (e.g. Figma) they jump straight to Claude or Cursor or v0 or whatever, where they're bounded by (and graded on) their ability to manipulate a bot, rather than their actual skillset.
It's not about vibe coding the best, it's about delivering value the fastest. And the value in this case is bringing clarity to the conversation when there are a lot of unknowns. As the article notes, prototypes are great for that. It's much easier to poke and prod at a prototype and use it to explore the problem space than to sit in on a presentation where the slides are just heavily distilled versions of an engineering design doc that is dozens of pages long (and most people will just skim anyway).
As a result, many service companies are moving to product businesses.
will it also happen to developers in other countries? I've no idea.
When I tell this to others, they go in disbelief, but people don't care till their own job is gone.
AI is very good in creating a working website. We do not have the issue of internet explorer 6 anymore, frontend doesn't host security relevant code and it shouuld also not have the business logic. So its easy to let ai create a website which talks to a secure api endpoint.
Plenty of basic android apps are just the same thing: Some API Endpoint calling Frontend.
I know plenty of people i worked with, which I would replace tomorrow if i got the choice of more tokens vs. having to work with them, putting in a lot of effort in Code Review and not seeing real benefits.
If the industry thinks we don't want to grow anymore, than the market gets saturated -> issue 1. Issue 2 -> if the market is saturated and tokens costs a lot of money budget wise, you will have to reduce your team size to compensate for the budget -> further reduction of market capacity.
We know that AI is replacing jobs
I don't have hard data on it, but if someone is spending $1.5k a month on Claude (the limit of some folks I know) that shakes out to $18k a year. In my experience, I'd MUCH rather use the $18k of a few developer's Claude budget to hire another dev instead of trying to get a robot to do stuff.
AI is useful, but it's can't f*cking learn like a human can and improve at something.
And, honestly, if your job can be replaced by AI, then your job was more susceptible to offshoring/layoffs/any other for m of cost cutting already.
I would replace that person in a heart beat tbh.
And when I look back, i can list a handful of people of top of my head which I would replace for these tokens too.
I still believe in ensuring that contributors understand every line of shipped code. This creates a natural ceiling on the width of their domain; though AI can help them learn more quickly, everyone can't know everything about engineering.
AI does not increase co-ordination costs per se; it increases productivity. You can achieve the same result with fewer people using AI, so it should decrease co-ordination costs. Only when you add AI while keeping the organization constant should co-ordination costs increase.
Sustainability will be most stretched where the bottom line actually matters. Medical devices and data processing, commercial aviation operations, precision manufacturing, and such. Bullshitting only takes you so far until actually doing the thing and showing that with high confidence becomes the only criterion.
I think AI has a lot of value when used with thoughtful care. I think a lot of the conversations around it are really about frustrations with charlatans.
I’m curious where you’re running into limits there. Intuitively, it feels like stuff like change lead time and MTTR is rooted enough in “what is the customer experiencing” that it shouldn’t have been made outdated by AI. But what are you finding?
The author could have written a rather incisive 800 words on this if he'd really tried.
But I will not read 2500 words of redundant, repetitive slop. It's really bad writing.
There is no pacing or conclusion to speak of. It's sort of just a loose list alternating between upsides and downsides, punctuated by the usual bullshit list-y ad-copy summations:
The build cost collapsed, the alignment cost rose, the thinking time disappeared, and the productivity gains got captured by output volume rather than output quality.> the thinking time disappeared
It did not. The author decided against spending time thinking. Nobody is following him around whacking him on the head if he blocks off tomorrow morning to go heads down on something. At least, I don't think they are?
A tiny bit of project management discipline can go a long way.
If you read between the lines -- and note, the author himself probably did not intend that subtext, or was careful to avoid implying it -- you will realize that the likely natural equilibrium involves way, way less people.
I wrote in more detail in this comment: https://news.ycombinator.com/item?id=48040999 -- but the upshot is that in the future teams, or heck entire departments, will be replaced by 1 - 3 very senior ICs that do both strategic and hands-on work.
The driving force for that structure is the high coordination costs or large org, which TFA bemoans quite a few times but does not explicitly call out as THE main issue. I mean, if three teams can provide working solutions in record time, your problem is not aligning them, your problem is that you have two teams too many.
On the bright side, fewer people also means fewer organizational pathologies like bureaucracy, silos, duplication of work, turf-wars, empire building and politics in general. The downside is well, a jobacalypse, starting with juniors (future talent pipeline problem says what?) and middle-management.
This does assume very capable ICs who can do strategic as well as tactical hands-on work. This is where future job seekers want to be. They will be worth their weight in gold. Which will still be wayyyyyy less than the value they provide (or crudely, the salaries of the teams they replace.) They'll do 5x the work but won't get paid 5x as much, of course. what do you think this is, socialism?
I want to point out if the organizational model or your team's engineers are resistant to change, it doesn't matter how good of an engineer you are, or how good at proposal writing you are. With or without AI.
Before AI, trying to program even a simple thing was an exercise in frustration from rules that had only been put in place by programmers to protect their own jobs and make it as difficult as possible for a normal person to develop. Oh! You mixed tabs and spaces, now your code will not compile and you're stuck another day. Oh! You forgot a semicolon, now the code won't run, even though the software points out your missed semicolon and thus knows how to fix it.
AI takes care of all that bagage and now I and others can make fully functional software that solves real world problem for real people.
I've spent the last three weeks working out a spec and didn't even start the development process yet.
The idea that the syntax of the language would ever be a bottleneck sounds ridiculous to me.
The gate keeping allegations are also incomprehensible. The vast majority of developers are working on making their jobs easier. There wouldn't be an endless stream of new programming languages, libraries and frameworks, if there was a software guild that you needed approval from to work on software. Even if such a guild existed, it would get obsoleted by the competition.
There are cases of people maximizing their own job security by writing terrible and incomprehensible code, but most experienced developers have gotten bitten by their own cleverness and try to make their code as easy to understand and as accessible as possible.
Things like Java Server Faces and Java Enterprise Edition died out a long time ago. The XML craze is over. Roy Fielding style REST/HATEOAS is dead and everything is an HTTP API with OpenAPI docs nowadays. People understand by now that micro service architectures only make sense for organisational purposes but not for technical reasons. NoSQL also waned and everyone is basically putting their JSON into PostgreSQL if they need to store complex hierarchical data.
Why do you even care about irrelevant things like semicolons? Like, any reasonable editor gives you squiggly lines so you can't miss them, meanwhile in practice having a line delimiter helps disambiguate hairy expressions and produce more readable error messages. For me they are an imperceptible cost that I couldn't care less about.
If you talked about null pointers, which are basically a landmine in every line you've ever written, waiting for a chance to explode, maybe you'd have a point but even nullable pointers are an idea that is being relegated to the history books.
Great! Then you can pick any man of the street and show him some code, and he will understand the syntax intuitively and start coding? Dollar signs, semicolons, brackets and === and the difference between "" and ''. It's all self explanatory.
Driving a motorized vehicle was a highly specialized task in the beginning. You had to prime fuel, adjust carb needles, maybe tighten a chain after a day of driving. Manufacturers did all they could to make vehicles as easy as possible for the users, so that they can focus on actually driving, and not fighting against the machine. Look at where cars are today - anybody can drive without needing any skills relating to the machine. AI is doing the same for programming, which is great.
Now a common man can make software without learning thousands of different arbitrary rules.
A programming language is way easier than learning a foreign natural language. I believe the issue you struggle with is formal logic, not the syntax. Not everyone is trained to think formally (and some may find it arduous).
That’s on the level of complaining about having to learn music theory to play the piano, or to learn grammar to write a report. Or having to learn the road rules to drive a car on the street.
But I understand that the majority of hackers here think that spelling mistakes in a report means that somebody is bad at their job as a police officer. And I know that a good portion of hackers here think that a suspect should have all charges dismissed if a police officer has mixed tabs and spaces when typing his final report, or forgot a semicolon. Or used ' where he should have used ".
Indentation rule are about scoping, You need a delimiter to mark end of a statements, and quoting often have to do with value type and interpolation. They’re not merely visual markers. Messing them up is the equivalent of answering “400 miles” when asked “what’s the color of the sky?”.
ADDENDUM: Yep writing code is filing a form. And just like any form, it’s easy to validate basic errors like syntax and type of values. The hard thing is to validate what happens after the form is processed. i.e, the intent of filing that specific form.
No it isn't anymore. Because of AI coding agents, all that is taken care of by the robot, and people can focus on other stuff.
> Like if the question was “Suspect Height” and you put “green”.
And any person reading the report can still make sense of mistakes. Such as if the answers are switched between two lines. But hackers would demand that the suspect has to go free because of a typo.
Thankfully the AIs are not anything like rule-obsessing hackers, and will understand exactly what you mean even if you made a typo or mistake somewhere.
At the time where machines have finally started behaving more human, a huge number of people insist on wasting their precious time on earth by selecting to have robot minds.