The one glaring omission the jumps out to me is Holmström's theorem like effects of incentives.
There has been a huge shift towards metrics, to the point that managers are often forced to or at least commonly believe they are forced to game those metrics.
It is challenging to take strategic risks when you have to focus on metrics which cannot measure let alone value those choices.
I have found working with remote first natives that the narrowness of their knowledge is also very high. When you work in an office there is a some knowledge transfer happening having lunch with the guy in accounts or the women in the sales team. This non structured learning is missed in remote work.
lol. Apparently folks where I work don’t ever talk at lunch. I work remotely, and over the course of about 20 meetings I discovered that what everyone “knew” was wrong because none of them talked to each other and each assumed they knew how it worked. Each person knew their own piece and had an incorrect idea of everyone else’s. There were twenty different systems in twenty different heads. These people all shared an office and lunch space. I work remotely.
Any large scale engineering product where “who you’ve talked to” is how things “get done” is going to fail.
Really you are just outing yourself as a member of the political class: someone who believes feelings and opinions matter to the behavior of CPUs.
The way this position conforms to the interests of the capital class, and conflicts with those of the labor class, is a red flag.
It simultaneously and conveniently:
1. takes the heat off AI blowback
2. synergizes perfectly with "RTO" mandates (to the extent this needed synergy to become A Widespread Thing)
On that basis alone, I'll wait for further analysis.
Edit: to be clear I'm no anti-AI holdout, and I actually don't mind working from the office (which i do 4x a week). Just observing.
AI is not responsible for anything at the moment, except making existing senior developers reasonably more efficient for sleeve of tasks, but not the tasks that take the most time.
Saying "we don't need as many staff because AI" is an oft-repeated trope because it sounds like a reasonable excuse to fire people. It's nearly impossible to back up the claim with any measurable method, and investors will look aside on the mismanagement and/or ridiculously over-engineered/over-complicated custom tech stacks companies run if they say "AI" anywhere in their reports.
There is plenty of companies already firing people due because of AI, or at least, that's why they proclaim.
There is also plenty of freelance / artistry type people who would of had work before (i.e creating Halloween, mothers day, Easter, etc) promotional material which is now just outsourced to AI. You see some of the biggest companies on Earth posting AI stuff for special event, etc.
Most aren't at the stage of using full AI "art" for advertisements (except maybe Coke) but some of these companies would already have full time artists, which they've bypassed. Their jobs are not forever and eventually will get killed.
AI is absolutely directly responsible. Managers are literally asking whether people need AI credits or interns over summer. Most people are taking credits, and internships are getting cancelled left and right.
At the moment is the significant part, how long do we expect this to last now that AI is capable of generating novel ideas like solutions to Erdos conjectures?
The narrative "we don't need as many staff because of AI" is a labor disciplining device whether or not it is true. Remaining engineers, loaded up with more projects under the threat of layoffs and with no outside opportunities, work nights and weekends to get them done. Then it is literally true that we accomplished more with less. And in a sense it is even "because of AI." But not in the way that you're supposed to think they mean it.
> AI is not responsible for anything at the moment, except making existing senior developers reasonably more efficient for sleeve of tasks, but not the tasks that take the most time.
I disagree. AI is often depicted as autonomous agents YOLOing features, but they excel at pattern matching from free form text and examples, and execute feedback loops. This means that they are particularly apt at small maintenance tasks spread across the project following clear high level guidance.
This is your typical junior task, the kind of task that is plausibly very boring and repetitive that is validation-heavy until it stumbles upon an unexpected turn and forces a senior to step in.
Once you offload these tasks to an agent running on a background, what exactly is left for a junior to do?
Juniors can arguably lean on AI coding agents to tackle more complex and more extensive work, but the truth if the matter is that they lack the skill and tools to effectively address this sort of work. They can get things to build but they fail to get things to make sense or be maintainable.
It's hard to really compare when we(the pre-internet generation) were raised in an offline environment from the start.
At my current job we see "junior" devs with 3 years of GitHub contributions and fully in production personal projects. Those obviously learned through a different path that what most had 20 years ago, but they're definitely not an exception either, and I genuinely think there is an adaptation process that many are missing.
Perhaps everyone can't follow that path. but not everyone could follow the previous one either. We'll probably only know when the dust fully settles.
Intial folks in gold rush benefits. Later folks don't. It happens everywhere. This weak junior hiring was seen by everyone 10 years ago when our politicians were asking coal miners to learn to code.
I have worked with graduates joining remotely during the pandemic, like most graduates they also lacked the skill to work in a real environment, but we can teach them, it's easy. But during the AI boom, the people who could teach the graduates were let go, leaving only a handful of senior engineers that had to "increase their productivity" while also mentoring the juniors. Guess where people cut corners to keep their job longer?
Yeah, I've had (and seen) interns and new-hires do fantastically during COVID.
You get what you pay for / put the work in for. If you're just hiring them, saying "read these docs and ship some PRs", and ignoring them, it's not too hard to predict what'll happen.
Engage at least as much as you would in person (more likely more, because you don't have passive hints about struggles), and it works out fine.
I have been working since 2008, in that time the only periods my manager has been within a hundred miles of me has been between 2010-2013 and 2015-2017.
Even if I pretend for a moment that a generation that is younger than Google is somehow unable to collaborate online, remote work has been the mode of operation of most people even before COVID, the only question is whether they are sitting in traffic or not first.
Now tell me which executives ever did the math on how remote working makes juniors take longer to learn and then took hiring decisions based on that math. This all seems good in theory, but doesn’t seem to hold out in the real world if you’ve ever worked with higher ups in your life
In 2026 a Junior Engineer is just Claude Code with a bad UI, higher latency, and extra steps. Literally.
I wouldn't even considering hiring a junior engineer at this point. The ROI was already barely breakeven for any but the top of the top junior engineers as they are likely to move on before they are meaningfully contributing.
With AI in the mix the ROI for Junior Engineers is strongly negative for 2 reasons:
1. (obvious) I can just have Claude Code do the work a junior engineer would have done with faster turnaround and generally better results.
2. (less obvious) Junior engineers are going to just turn around and use Claude Code, so now I'm talking to an AI and playing the telephone game, and the Junior engineer isn't going to learn much if anything in the process.
People don't want to hear this, but it's true, especially the part about the junior developer just using Claude Code themselves.
We may still be too early on the curve of AI usage for AI to be the major driver of the labor market changes, but we also have no clue about what to do about it.
Often the conversation puts "Using AI ourself" at odds with "Delegating to a junior developer", but the junior developer is going to be using AI just like the rest of us further bringing into question the value of junior developers (and eventually senior engineers).
What really is the next step? (Rhetorical question since nobody knows)
If remote work is to blame, wouldn't that also indicate that you couldn't outsource part of your work to lower salaried countries. A lot of companies have their senior staff at the home office, then outsource the more trivial tasks to e.g. India.
Managing people is hard, managing remote people makes it hard to hide that you're a poor manager. If you want to allow for remote work, you need to invest in management training. You should do that for onsite as well, but it will be less visible that you if you skip that step.
What if unrealistic economical growth is to blame for weak junior hiring?
Or how about we make remote work mandatory where possible so the economy lets people live their lives. Getting back unpaid time from commutes and being able to reorganize work time freely makes a huge difference.
If my location dictates the type of employees your organisation needs / doesn't need, then yeah, you pretty much over-hired to begin with and just lack accountability. Hence trying to blame it on everyone and everything else except yourself. Respectfully.
In my anecdotal experience in a FAANG, weak junior hiring started during the hiring freezes in mid 2022, and was made worse by the layoff cycles that began soon after. Once you know headcount is going to be extremely tight indefinitely, you want to use your precious few slots to hire someone that can deliver value pretty quickly, rather than take years to coach up.
It personally seems hard to connect that to remote work as that had been going for 2 years and in between was the largest hiring burst we'd done, which included many junior folks. Though admittedly I'm biased as a remote worker.
It's telling how rhetoric and conjecture are now normalized in company and business culture. When we were at the peak of remote work, companies were reporting record high revenues.
I got significantly more work accomplished when I worked remotely. There was tons of work I could not do from the office as people needed the systems I need to work on and could only be done out of hours/remotely.
When in the office I got a lot of people complaining/pissed I was leaving early because I got there an hour or more before everyone so I could get more work done/do the work on systems they needed. The only thing I got while in the office was constant interruptions for things a junior could have handled. Meetings was a bad word, never allowed, so there was really no reason for me to be there constantly (I could have done most of my job remotely and gone into the office once or twice a week)
I hired a junior I was eager to mentor/train to replace me, they proceeded to throw endless things I did at them expecting them to fill the multiple hats I was filling (to the point they pushed them to work on very dangerous equipment and got themselves hurt)
Dammed if you do, dammed if you don't
I absolutely loved the work I did. I GTFO of that misery that was only miserable/got me crucified due to the stupid shit people made up in their head instead of the actual work I did
Companies used to invest in building up their employees and even in retaining them by giving competitive pay. I can hire any junior developer and train them to be a better developer, unless they themselves do not want to be a better developer, I cant fight lack of motivation.
A lot of senior developer roles list things that make it sound like senior devs are supposed to mentor other devs but they never seem to do so.
Junior hiring was clearly depressed thanks to COVID, but AI has absolutely made that depression worse. We know this because it's what the tech leaders keep saying. They can't fucking shut up about it.
It's as though the only idea any tech CEO has had in the last year is "what if we gave our best engineer 1000 agents and a case of Red Bull and fired everyone else?" Historically you had to hire junior engineers, because you needed the help. Now there's a [theoretical, purely imaginary] world where you don't, because agents will magically do that work for you. Nobody is losing sleep over the effect this has on talent as a whole, because that's a problem for someone else in the future.
It's mismanagement, a prevalence of PE pushing profit margins as thin as possible, and the inevitable feeling of an oncoming recession. Mismanagement and PE both push to prioritizing short term gains (something you can use to justify your position/investment today) at the cost of long term profitability. No one is getting a bonus for having a great quarter in 2046 when your new project has turned you into a trillion dollar company. Executives tend to be very gullible and believe the department head that will claim it wasn't R&D but the new slick UX that 10x'd the company.
Add to that the economy, especially after the disastrous Trump administration, which we can all plainly see as an oncoming train heading straight towards us, and even those who would optimistically advocate for long term budgeting in good times are in baton down the hatches mode.
Hiring juniors is an excellent long term strategy that takes time to pay off - you're much better off having a mix of labor that can mix bold initiative and raw enthusiasm with prudence and planning - and those junior devs today will turn into highly skilled professionals with a deep understanding of your platform in half a decade or so. But when times are lean that's difficult to justify.
I wouldn't shift all the blame away from AI though, this isn't a singular cause thing - working for a PE owned firm we're now on the hook for 200/mo anthropic seats owing to our overlords making a horrible deal. The current brand of AI is a rent-seeking technology that's pulling funds out of the working areas of the economy to fund its insane R&D concepts while more traditional AI applications that have proven utility are languishing,
97 comments
[ 4.1 ms ] story [ 105 ms ] threadhttps://papers.ssrn.com/sol3/papers.cfm?abstract_id=6787638
Unclear if it's been peer reviewed. The abstract looks fairly convincing. But it is argued against by the majority of research on this topic.
There has been a huge shift towards metrics, to the point that managers are often forced to or at least commonly believe they are forced to game those metrics.
It is challenging to take strategic risks when you have to focus on metrics which cannot measure let alone value those choices.
Of course if everyone is working remotely via email this isn't going to happen.
I've had the same problem in person too (silos, no one talking) so I think it's more about structure than remote/in-person.
Any large scale engineering product where “who you’ve talked to” is how things “get done” is going to fail.
Really you are just outing yourself as a member of the political class: someone who believes feelings and opinions matter to the behavior of CPUs.
It simultaneously and conveniently: 1. takes the heat off AI blowback 2. synergizes perfectly with "RTO" mandates (to the extent this needed synergy to become A Widespread Thing)
On that basis alone, I'll wait for further analysis.
Edit: to be clear I'm no anti-AI holdout, and I actually don't mind working from the office (which i do 4x a week). Just observing.
Saying "we don't need as many staff because AI" is an oft-repeated trope because it sounds like a reasonable excuse to fire people. It's nearly impossible to back up the claim with any measurable method, and investors will look aside on the mismanagement and/or ridiculously over-engineered/over-complicated custom tech stacks companies run if they say "AI" anywhere in their reports.
There is also plenty of freelance / artistry type people who would of had work before (i.e creating Halloween, mothers day, Easter, etc) promotional material which is now just outsourced to AI. You see some of the biggest companies on Earth posting AI stuff for special event, etc.
Most aren't at the stage of using full AI "art" for advertisements (except maybe Coke) but some of these companies would already have full time artists, which they've bypassed. Their jobs are not forever and eventually will get killed.
I disagree. AI is often depicted as autonomous agents YOLOing features, but they excel at pattern matching from free form text and examples, and execute feedback loops. This means that they are particularly apt at small maintenance tasks spread across the project following clear high level guidance.
This is your typical junior task, the kind of task that is plausibly very boring and repetitive that is validation-heavy until it stumbles upon an unexpected turn and forces a senior to step in.
Once you offload these tasks to an agent running on a background, what exactly is left for a junior to do?
Juniors can arguably lean on AI coding agents to tackle more complex and more extensive work, but the truth if the matter is that they lack the skill and tools to effectively address this sort of work. They can get things to build but they fail to get things to make sense or be maintainable.
So what is a junior dev to do?
At my current job we see "junior" devs with 3 years of GitHub contributions and fully in production personal projects. Those obviously learned through a different path that what most had 20 years ago, but they're definitely not an exception either, and I genuinely think there is an adaptation process that many are missing.
Perhaps everyone can't follow that path. but not everyone could follow the previous one either. We'll probably only know when the dust fully settles.
You get what you pay for / put the work in for. If you're just hiring them, saying "read these docs and ship some PRs", and ignoring them, it's not too hard to predict what'll happen.
Engage at least as much as you would in person (more likely more, because you don't have passive hints about struggles), and it works out fine.
Yeah.
Even if I pretend for a moment that a generation that is younger than Google is somehow unable to collaborate online, remote work has been the mode of operation of most people even before COVID, the only question is whether they are sitting in traffic or not first.
I wouldn't even considering hiring a junior engineer at this point. The ROI was already barely breakeven for any but the top of the top junior engineers as they are likely to move on before they are meaningfully contributing.
With AI in the mix the ROI for Junior Engineers is strongly negative for 2 reasons:
1. (obvious) I can just have Claude Code do the work a junior engineer would have done with faster turnaround and generally better results.
2. (less obvious) Junior engineers are going to just turn around and use Claude Code, so now I'm talking to an AI and playing the telephone game, and the Junior engineer isn't going to learn much if anything in the process.
We may still be too early on the curve of AI usage for AI to be the major driver of the labor market changes, but we also have no clue about what to do about it.
Often the conversation puts "Using AI ourself" at odds with "Delegating to a junior developer", but the junior developer is going to be using AI just like the rest of us further bringing into question the value of junior developers (and eventually senior engineers).
What really is the next step? (Rhetorical question since nobody knows)
Managing people is hard, managing remote people makes it hard to hide that you're a poor manager. If you want to allow for remote work, you need to invest in management training. You should do that for onsite as well, but it will be less visible that you if you skip that step.
What if unrealistic economical growth is to blame for weak junior hiring?
It personally seems hard to connect that to remote work as that had been going for 2 years and in between was the largest hiring burst we'd done, which included many junior folks. Though admittedly I'm biased as a remote worker.
But let's not pretend reality enters the decision-making of the large tech company at any point, for any kind of decision.
When in the office I got a lot of people complaining/pissed I was leaving early because I got there an hour or more before everyone so I could get more work done/do the work on systems they needed. The only thing I got while in the office was constant interruptions for things a junior could have handled. Meetings was a bad word, never allowed, so there was really no reason for me to be there constantly (I could have done most of my job remotely and gone into the office once or twice a week)
I hired a junior I was eager to mentor/train to replace me, they proceeded to throw endless things I did at them expecting them to fill the multiple hats I was filling (to the point they pushed them to work on very dangerous equipment and got themselves hurt)
Dammed if you do, dammed if you don't
I absolutely loved the work I did. I GTFO of that misery that was only miserable/got me crucified due to the stupid shit people made up in their head instead of the actual work I did
A lot of senior developer roles list things that make it sound like senior devs are supposed to mentor other devs but they never seem to do so.
It's as though the only idea any tech CEO has had in the last year is "what if we gave our best engineer 1000 agents and a case of Red Bull and fired everyone else?" Historically you had to hire junior engineers, because you needed the help. Now there's a [theoretical, purely imaginary] world where you don't, because agents will magically do that work for you. Nobody is losing sleep over the effect this has on talent as a whole, because that's a problem for someone else in the future.
Add to that the economy, especially after the disastrous Trump administration, which we can all plainly see as an oncoming train heading straight towards us, and even those who would optimistically advocate for long term budgeting in good times are in baton down the hatches mode.
Hiring juniors is an excellent long term strategy that takes time to pay off - you're much better off having a mix of labor that can mix bold initiative and raw enthusiasm with prudence and planning - and those junior devs today will turn into highly skilled professionals with a deep understanding of your platform in half a decade or so. But when times are lean that's difficult to justify.
I wouldn't shift all the blame away from AI though, this isn't a singular cause thing - working for a PE owned firm we're now on the hook for 200/mo anthropic seats owing to our overlords making a horrible deal. The current brand of AI is a rent-seeking technology that's pulling funds out of the working areas of the economy to fund its insane R&D concepts while more traditional AI applications that have proven utility are languishing,