The Information: Microsoft Engineers Forced To Dig Their Own AI Graves 71
Longtime Slashdot reader theodp writes: In what reads a bit like a Sopranos plot, The Information suggests some of those in the recent batch of terminated Microsoft engineers may have in effect been forced to dig their own AI graves.
The (paywalled) story begins: "Jeff Hulse, a Microsoft vice president who oversees roughly 400 software engineers, told the team in recent months to use the company's artificial intelligence chatbot, powered by OpenAI, to generate half the computer code they write, according to a person who heard the remarks. That would represent an increase from the 20% to 30% of code AI currently produces at the company, and shows how rapidly Microsoft is moving to incorporate such technology. Then on Tuesday, Microsoft laid off more than a dozen engineers on Hulse 's team as part of a broader layoff of 6,000 people across the company that appeared to hit engineers harder than other types of roles, this person said."
The report comes as tech company CEOs have taken to boasting in earnings calls, tech conferences, and public statements that their AI is responsible for an ever-increasing share of the code written at their organizations. Microsoft's recent job cuts hit coders the hardest. So how much credence should one place on CEOs' claims of AI programming productivity gains -- which researchers have struggled to measure for 50+ years -- if engineers are forced to increase their use of AI, boosting the numbers their far-removed-from-programming CEOs are presenting to Wall Street?
The (paywalled) story begins: "Jeff Hulse, a Microsoft vice president who oversees roughly 400 software engineers, told the team in recent months to use the company's artificial intelligence chatbot, powered by OpenAI, to generate half the computer code they write, according to a person who heard the remarks. That would represent an increase from the 20% to 30% of code AI currently produces at the company, and shows how rapidly Microsoft is moving to incorporate such technology. Then on Tuesday, Microsoft laid off more than a dozen engineers on Hulse 's team as part of a broader layoff of 6,000 people across the company that appeared to hit engineers harder than other types of roles, this person said."
The report comes as tech company CEOs have taken to boasting in earnings calls, tech conferences, and public statements that their AI is responsible for an ever-increasing share of the code written at their organizations. Microsoft's recent job cuts hit coders the hardest. So how much credence should one place on CEOs' claims of AI programming productivity gains -- which researchers have struggled to measure for 50+ years -- if engineers are forced to increase their use of AI, boosting the numbers their far-removed-from-programming CEOs are presenting to Wall Street?
Re:not so different (Score:5, Funny)
Re: (Score:2)
Re: (Score:2)
The H1Bs aren't better. They are just paid half as much and unable to compete fairly in the marketplace because they're slave-labor tied to the H1B visa and have to go home if their employer ends it. Companies like MS rotate the H1Bs every 2-3 years to artificially keep the pay scale low.
Now go fuck yourself, Buck-Yar the Crossburning American-Hating KKK Fucking Shitbag.
Re: (Score:2)
Will require some new/tweaked visa fraud checks to avoid companies specializing in getting people H1B who then leave immediately to work elsewhere, but is very doable.
Re: (Score:2)
Re: not so different (Score:1)
I don't see the issue:
MS Exec has 400 employees under him.
Those 400 workers used AI to write 20-30% of their code.
Exec tells them to increase usage of AI to write up to 50% of new code.
MS lays off 6,000 employees across organization, this MS Exec loses 12 workers.
So a 3% staff reduction is attributed to increasing the use of AI to generate new code? That makes no sense, and there's no logical way to tie the headcount reduction to increased AI use.
MS has about a quarter million employees, they cut about 2% o
It's a good day. (Score:5, Funny)
I do enjoy a nice bonfire. Especially when it's the self-immolation of MS.
Re: (Score:3)
Indeed. The faster they are gone, the better for everyone.
Re: (Score:1)
Their killer feature is still Email plus Calendar.
if only Google had strategic direction (Score:3, Insightful)
There are so many things Google could have killed off by putting the slightest bit of effort into the situation. Like Okta, why does that company even exist? It's a $20 billion company that could have easily been strangled in the cradle by Google if Google had offered and marketed a SSO solution. They already had the technology but it was not a strategic priority. Microsoft came along later with its offering (which to be frank has eaten away at Okta's share, because most companies need to have MSFT for one
Re: (Score:3)
Like Okta, why does that company even exist?
Good question. Very good question [techcrunch.com]
You're not wrong (Score:2)
I mean, their two competitors were Lotus Notes and Novell Groupwise. Now, it's whatever Google hasn't killed as a product this week and... Zoho?
It won't play out that way. (Score:5, Insightful)
This very-visible move is just another marketing technique. "See how many expensive programmers we were able to let go by using our products? Just imagine how many expensive programmers YOU can let go by using our products! Send your money our way!"
And it will certainly work in the short run. Many will buy into the hype, and Microsoft will profit greatly from it. More than enough to quietly hire from the now-sizable pool of unemployed programmers once they are needed to work on the next important thing.
Microsoft is too rich to suffer real consequences of anything unwise they may do.
Re: (Score:2)
I'd imagine that the next important thing will be cleaning up all the security and performance issues caused by sloppy AI enhanced "vibe" coding, which seems to be coming all the rage in 2025.
Re: (Score:2)
I'd imagine that the next important thing will be cleaning up all the security and performance issues caused by sloppy AI enhanced "vibe" coding, which seems to be coming all the rage in 2025.
It it's becoming all the rage in 2025, maybe it will be causing all the rage in '26 or '27?
Re: (Score:2)
Re: (Score:2)
> "See how many expensive programmers we were able to let go by using our products? ..."
You see, I wonder... Microsoft obviously does have some talented developers somewhere, but it seems the rank and file are, as far as it's possible to be, just code monkeys. I draw this conclusion from the utter shit that Microsoft produces - it wasn't developed by a high-functioning, talented team of people, it was top-down designed, then cut down to the bare minimum by management and then built by people who either
False sense of economy (Score:2)
I wonder if the MSFT executive will be held accountable when more problems are introduced by developers who do not fully understand what the AI-generated code actually does.
Re:False sense of economy (Score:5, Insightful)
Microsoft? If they ever become accountable for all the crappy tech and all the bad design they are responsible for, the company is dead.
Re: (Score:2)
Re: (Score:2)
The largest economy in the world is dead
Not yet, the damage from the rule of Xi will manifest itself fully in the coming decades.
Re: (Score:2)
China may catch up with the US in decades... but even that is just a maybe.
Re: (Score:2)
Given the faster-than-light orbanization of the US, it is hard to take optimists seriously.
I have yet to see a society that glorifies corruption, prays to conspiracies, shits on basic democratic principles and isolates that simultaneously does well otherwise.
It won't be a case of China catching up, but rather one of the US racing against China.
And it will be a short one.
That's probably not going to happen (Score:3)
The first is that the AI works and it's successfully replaces the workers who got fired.
The second is that the AI doesn't work. However we are purposefully going into a recession, possibly even a depression. And we are unlikely to ever emerge from that because the things we need to do to emerge from recession we just do not want to do.
This means that if the AI doesn't work anyone who is still left holding a job is going to be forced to work double shifts to make up for
Re: (Score:2)
I don't know. I generally think that we're all replaceable, and obviously we mostly are. That said, I feel like those of us on the slightly above average and higher massively overestimate how much slightly below average to below average make things kinda hard to get done. What I mean is that it seems to be incredibly hard to hire and keep decent employees. Sure, there are people who will show up and do the most defined tasks - the sorts we have been automating long before AI. But there seem to be a real sho
Re: (Score:2)
We don't hold our ruling class accountable (Score:3, Insightful)
But at least those 14 trans girls playing field hockey aren't going to be able to steal your scholarships or som
Re: (Score:2)
You can hold them accountable all you want, but where is the enforcement mechanism other than voting every 2-6 years? (Primary challenge here, primary challenge there) Besides, people forget, and it doesn't help that they're apathetic. There is no referendum and recall for elected officials in the federal government. 2-6 Years a a long time where there is no mechanism to oust an elected official who is a loose cannon other than impeachment, and a lot of damage (as well as a lot of good in some cases) can
Re: (Score:1)
I know it's cool to hate on capitalism and billionaires at the moment but simplistically in a capitalist society people can acquire vast amounts of money by producing something that other people get value from more efficiently than their competition.
The more capital those people acquire, the more they can put to good use doing other stuff. If they succeed they make more, if not they lose it.
Contra
Re: (Score:2)
And now to the point of your argument, which basically sucks, and is totally wrong:
>I don't have a problem with successful capitalists getting richer and controlling more money
I do. They are stepping over the corpses, literally, of consumers and exploiting their workers. How else do you THINK they become billionaires? Being good?
Re: (Score:2)
you say you're left leaning but then you go on a right wing diatribe
no one hates on capitalists or capitalisms, it's just not what we have, we have pseudo-capitalism
It cannot be capitalism when 85% of all of the capital is controlled by 15% of the people, that only leaves 15% of all capital for the rest of the 85% of the people. That's not enough capital, 85% of us are severely undercapitalized. The end result is 85% of us are unable to participate in capitalisms.
In a corrupt economy, the upper class never
no way 20-30% (Score:4, Interesting)
I do not believe for a second that 20-30% of deployed code at MS has been written by LLMs. I believe that a giant volume of it has been *generated*, reviewed, found not usable, and discarded.
I wish some less credulous tech press person would actually press CEOs on idiotic claims like this.
Re: (Score:2)
Maybe it goes like this: AI creates 1000% of all code. 980% are found to be complete crap and 20% are found to be usable small fragments, by pure random chance. Well, "usable" for Microsoft-levels of "quality", that is.
Does it matter? (Score:3)
It's a win-win. If the AI works they don't have to hire employees and if it doesn't the few employees who are left just have to work 80 to 90 hours a week cranking out code or they get fired and replaced with one of the hundreds of unemployed programmers.
One way or another we are running out of useful work for people to do. This happens during the last two
Re: (Score:2)
" It took decades of technological advancement and two world wars"....the first of which happened concurrently with the most devastating global pandemic in 500 years
Re:no way 20-30% (Score:5, Interesting)
I do not believe for a second that 20-30% of deployed code at MS has been written by LLMs.
I do. There's a heck of a lot of code that is obvious, and can be filled in by autocomplete. You start typing a getter and it spits out the most common implementation. Now you have code that was written by LLM. Whether it saves any actual time or not compared with copying and pasting a similar method and tweaking it is a different question, of course.
Thus far, my experience is that it sometimes saves a decent amount of time. And it sometimes takes so much work that it would have been an order of magnitude quicker to do it yourself. And there's no obvious pattern when it comes to determining which is which. So you average out at a break-even level or even a slight loss of efficiency. But hey, you used AI. :-D
Re: (Score:2)
Yeah I agree as well. I haven't had to do much MS code for a while, but wow, they just generally have the crappiest architecture design I've ever encountered. I mean it is the place where the lament 'architectural astronaut' was born.
When I was younger, I tried to understand what they were doing with their various driver models, franken-unix style file system, bizarre layer-apon-layer of UI 'object' models, and that nightmare that is the registry. It's just all horrible, and I genuinely think many of the or
Re: (Score:2)
>When I do MS work, I do find myself just copying and pasting big chunks of code because life is too short to try to understand if there is any sort of coherent design going on (and these days I assume there isn't). If AI code paste can help untangle that, then good for them.
No, in the end this will be copying bigger chunks of totally incomprehensible code that "works" all over the place. It won't help untangle that, it will make it worse.
Re: (Score:2)
Good news- since the CEO just misled you, you can sue him for breach of fiduciary responsibility.
But the fact is- he's probably telling the truth, and your ego is really struggling to come to grips with that.
Re: (Score:2)
I do believe it. There's endless blobs of XAML, CSS, boilerplate HTML, that are currently squeezed out by disposable H1Bs.
Re: (Score:2)
I do not believe for a second that 20-30% of deployed code at MS has been written by LLMs.
I spoke with a friend working at Microsoft. He doesn't believe the 20-30% number either.
Something folks need to realize (Score:2)
If it works that's great for Microsoft they can replace employees.
If it doesn't work that's great for microsoft, they fire the employees they couldn't replace and force everyone to work double hours. The alternative is unemployment and eventual homelessness.
Sure human beings can't work 60 to 80 hours a week indefinitely but there's plenty more where that came from after you burn out the last batch.
A complete lack of competition, out of control we
story seems weak (Score:3)
My read on this is that Microsoft wanted to get a good measurement of the difference in productivity of the software devs when they were using AI as opposed to not. Useful to know for sure. Apparently they liked what AI could do and that may have initiated the layoff of 12 out of the 400 software devs on this guy's team.
Or not, the story seems pretty weak. Also "forced to dig their own AI graves" seems like a stretch. If these guys were writing AI software that would then turn around and put them out of work? That would have the appearance of grave digging. Merely using AI technology to do some work does not.
Re: (Score:2)
My read on this is that Microsoft wanted to get a good measurement of the difference in productivity of the software devs when they were using AI as opposed to not. [...] "forced to dig their own AI graves" seems like a stretch. If these guys were writing AI software that would then turn around and put them out of work? That would have the appearance of grave digging. Merely using AI technology to do some work does not.
Your comment lacks internal consistency. If your first assumption is correct, then the last sentence I quoted here makes no sense, because they are assisting by testing. QA is an important part of software development, and you are saying they are doing it!
Re: (Score:2)
I don't see how measuring productivity is 'QA', and "Merely using AI technology to do some work" is not QA.
The article seems to be trying to imply that the mere act of using AI for software development at all is equivalent to working yourself out of a job. That's a stretch. But I am hearing that the AI companies are enthusiastically using their own products to help generate and improve code for AI. Does this mean they get to lay off their programmers? Maybe it simply means they get a lot more done in a lot
Not such a great idea (Score:2)
Laying off the people with actual intelligence at a company doesn't seem like a prescription for success to me. Eventually they will end up at firms and startups that will do things bigger and better than Microsoft has. And Microsoft is almost a decade and a half past its glory days. "AI" will not bring them back. Hard work and a touch of sanity will.
Re: (Score:2)
Eventually they will end up at firms and startups that will do things bigger and better than Microsoft has.
And Micro$oft will buy them up and bury their tech like they did the last 100 start-ups that did things better that M$.
Re: (Score:2)
That is why you need startup founders, employees, and investors who care more about changing the world, helping people, and making cool stuff than putting millions (or billions) of dollars into their bank accounts in three years instead of ten.
Re: (Score:2)
That is why you need startup founders, employees, and investors who care more about changing the world, helping people, and making cool stuff than putting millions (or billions) of dollars into their bank accounts in three years instead of ten.
Once they go IPO, shareholder value trumps everything else
Monopolies. (Score:5, Insightful)
In a competitive economy, Microsoft couldn't have made $171,008,000,000 gross profit last year because that would have had to have been invested in development to keep up with the competition.
Similarly if AI is taking workload off developers, you start more development projects, rather than cutting costs and increasing profits.
Not only that (Score:2)
Where my kid is one private equity firm owns 80% of the places they can possibly work. You can imagine what that does to you competition and wages.
But good luck getting voters to care. They're too busy with moral panics
It's the thought that counts? (Score:4, Funny)
Clippy: I see you're digging your own grave. Would you like some help?
Copilot: Clippy, I see you're trying to assist. Would you like some help?
Clippy: No. Fuck off.
GitHub Copilot Coding Agent at Work (Score:5, Interesting)
Meet your new AI co-worker, kids! My new hobby: watching AI slowly drive Microsoft employees insane [reddit.com]
Interesting priorities (Score:3)
I wrote software for Microsoft for a long time - I joined as an SDE2 and left as a principal engineer.
Every team that I was on would routinely (like, weekly) close bugs and feature requests as "won't fix" not because the bugs weren't real or the features weren't a good idea, but because we just didn't have enough people to do everything that we would have liked to do.
So now that the company claims that AI is writing 30% of the code, are they raising the quality bar, or adding more features that customers want? Nope, they're firing thousands of engineers and distributing more profit to shareholders.
I'm not at all surprised, but I'm still a little bit disappointed.
Re: Interesting priorities (Score:2)
I have worked mostly in big tech as well, though not MS. The feature rejections are a fact of life. Closing legitimate bugs slightly less so. That's why many of my decades old Mozilla bug reports still remain open today, with a priority set low enough that no product manager will ever see them in their report again, but still there for everyone who runs into the same issues in 2025 to know that they weren't the first.
Re:Interesting priorities (Score:4, Interesting)
I worked as a programmer in QA (and got laid off because supposedly the 3 engineers in India or 4 in China can do it better - they can't - I support the product stil through contracting), and I've found so many easy exploits. The AI code is even worse, it often doesn't do bounds checking or allows code injections (that is my baby, been injecting system exploits for years). Some of that can be fixed by language used, but the product I've supported for years uses C++, which is a black hole of exploitability. I'm sure it will get better as time goes on, but it isn't good now.
I'm not against AI writing code, I've learned a lot from AI generated code, but it needs to learn a lot from human generated code, too, lol.
Re: (Score:2)
I spent practically my whole career as a maintenance programmer. Bugs that escaped to the field came back to me to find and fix. Shifting the bulk of coding to AI scares the willies out of me. Not because it would make my job obsolete, but because it would double or triple my workload with some of the most esoteric, inscrutable, and just plain bad code ever seen by mortal man.
Look, we've all pulled up old code and said to ourselves, "How the hell did this ever work in the first place?" AI code will be wor
LLM collapse anyone? (Score:2)
It will come around to bite them. (Score:5, Interesting)
And quite a few people at Microsoft know it. Some at the company are trying so hard to make AI happen, but there are other segments (the major cash cows) are starting to realize that Copilot is Clippy and Cortana 2.0 and nobody is buying tickets for the sequel. Office (quietly) added options to remove it (you know, for compliance reasons, cough) and introduced the "classic plan" for Office.
Search "Copilot Key" and the biggest result is how to remap it to whatever you want.
The Azure group reduced their overall hardware rollout in this space and they did it because the internal demand wasn't showing up.
They open sourced all the bits around Copilot integration in VSCode because "AI is so core to modern development experiences it makes the most sense to integrate this..." They actually couldn't justify the expense in maintaining it otherwise and it undercut all the insanely overvalued code forks. They also open sourced the WSL bits, but that was just long overdue anyway.
So, when this is all over, they'll have to silently hire back engineers again and it will cost them a fortune to find because the pipeline will be drier than ever. I use Windows as my daily driver, but even the most die-hard of MS fans knows this was a bad move. They would be in more trouble, save for the fact that everybody else is doubling-down on AI whatever and making equally dumb mistakes.
A good engineer is better than a good AI (Score:2)
Good engineers need not fear anything, they will only see an increase in productivity by using AI for mundane coding tasks. Poor performing engineers, and MS only dumped 12 of 400, can expect to be replaced by a very small shell script.
This is the way it's supposed to work. Anything that helps a company drop dead weight is good for the bottom line.
I wouldn't be surprised that... (Score:1)
Then on Tuesday, Microsoft laid off more than a dozen engineers on Hulse 's team as part of a broader layoff of 6,000 people across the company that appeared to hit engineers harder than other types of roles, this person said."
I wouldn't be surprised that the majority of Microsoft employees are considered 'engineers', so that a broad layoff of 6,000 workers "hit engineers harder than other types of roles"...
Generating code is not what counts... (Score:1)
Generating quality, efficient, and bug-free code is what counts. Which AI is incredibly bad at.