Is A.I. Replacing Tech Workers or Providing an Excuse for Job Cuts?

https://www.nytimes.com/2026/06/01/technology/ai-tech-job-cuts.html?smid=nytcore-ios-share

But in more than a few cases, the recent layoffs have coincided with other business issues. Wall Street loves an A.I. story right now. That, analysts and economists say, has offered a smoke screen for companies looking to beef up profits or patch over old mistakes.

Cutting jobs to make way for A.I. is “a nice excuse, but some of these aren’t necessarily the best, most well-run companies,” said Mark Mahaney, an analyst at the investment bank Evercore. “They may have overhired, or they may be losing market share. There may be other issues.”


“All these cuts are happening, and there are record profits,” said Ava Sazanami, who worked for Meta from 2022 to 2025. A.I. “is actually not costing any less money,” she added. “It is an excuse to some extent.”

Without reading the article the answer is yes.

There was a crapload of funding made available to businesses during COVID to hire staff. That funding is gone and a lot of workers just don’t provide any actual value and much of their work can be automated away or simply reassigned to somebody else who AI tools have made more productive.

I know people who exist on both sides of this issue, it sucks, nobody is happy, but it’s the reality of the situation.

Excuse, has been shown many times, now include the sticker shock for companies that did actually go all in on AI, and now they are looking to hire people back because AI is costing them 10X what the workers did...

Lakados said:

Without reading the article the answer is yes.

There was a crapload of funding made available to businesses during COVID to hire staff. That funding is gone and a lot of workers just don’t provide any actual value and much of their work can be automated away or simply reassigned to somebody else who AI tools have made more productive.

I know people who exist on both sides of this issue, it sucks, nobody is happy, but it’s the reality of the situation.

Click to expand...

I always hated it when I got laid off from my tech job, always during a business downturn or after the "tecyh bubble" burst in 2000-2001.

There's a balance to be found.

• You can't lay-off too many people; your revenue stream will dry up.
• You can't hire too many people; It's impossible to direct them.
• You can't replace people with AI; AI cannot come up with new ideas.
• AI is prohibitively expensive to maintain, and it makes careless errors that are hard to spot.

At this point, AI = "quick profit" to many CEO's, and they've been sold a lie. They're starting to realize it. Now they're trying to make it seem like they won the AI race by cutting jobs and saying "We've modernized with AI agents", then going to the remaining workers and saying "Make AI work!!!"

Either the workers will make AI work, the company will rehire it's laid-off talent, or it will die. Capitalism is already in transition, and there's no telling what it will look like when it's over.

MrGuvernment said:

Excuse, has been shown many times, now include the sticker shock for companies that did actually go all in on AI, and now they are looking to hire people back because AI is costing them 10X what the workers did...

Click to expand...

Claude Code costs $200/month. A software engineer probably averages $16,000+/month in total compensation. It accelerates what one person can do in many positions by 20-300%+.
People need to be honest with what is happening - AI is replacing jobs.
I've seen firsthand what it can do. You don't need analysts anymore. You don't need secretaries. You don't need software engineers. You don't need finance people. You don't need financial accountants. You don't need anyone who did anything in Excel or PowerPoint. You don't need as many middle managers.

sleepeeg3 said:

Claude Code costs $200/month. A software engineer probably averages $16,000+/month in total compensation. It accelerates what one person can do in many positions by 20-300%+.
People need to be honest with what is happening - AI is replacing jobs.
I've seen firsthand what it can do. You don't need analysts anymore. You don't need secretaries. You don't need software engineers. You don't need finance people. You don't need financial accountants. You don't need anyone who did anything in Excel or PowerPoint. You don't need as many middle managers.

Click to expand...

And it's this sort of mentality that will lead a company to ruin. AI is great, but for those of us that have actually used it, you have to hold it's hand. Otherwise, it will, at best, not work and, at worst, destroy everything you've worked for. Ask me how I know...

sleepeeg3 said:

Claude Code costs $200/month. A software engineer probably averages $16,000+/month in total compensation. It accelerates what one person can do in many positions by 20-300%+.
People need to be honest with what is happening - AI is replacing jobs.
I've seen firsthand what it can do. You don't need analysts anymore. You don't need secretaries. You don't need software engineers. You don't need finance people. You don't need financial accountants. You don't need anyone who did anything in Excel or PowerPoint. You don't need as many middle managers.

Click to expand...

$200 and you know how fast you can burn through tokens? Especially with how Anthropic has changed various things like their time in memory?

I am not saying AI is not a useful tool, it is very useful in the hands of those who know how to use it, but as I noted, excuse, as there are reports now of companies literally going WTF....why do we have a bill for $$$$$$ when it was only supposed to be $200 a month/per seat?

And all these roles are absolutely still required - because "AI" can not be trusted... ask me how I know.....

You don't need software engineers. You don't need finance people. You don't need financial accountants.

Click to expand...

MrGuvernment said:

$200 and you know how fast you can burn through tokens? Especially with how Anthropic has changed various things like their time in memory?

I am not saying AI is not a useful tool, it is very useful in the hands of those who know how to use it, but as I noted, excuse, as there are reports now of companies literally going WTF....why do we have a bill for $$$$$$ when it was only supposed to be $200 a month/per seat?

And all these roles are absolutely still required - because "AI" can not be trusted... ask me how I know.....

Click to expand...

I do know how quickly you can burn through tokens. It is not equivalent to one or several employees that it can replace, for actual work.
There are also other AI tools out there that cost a fraction as much. Copilot is $30/month and now supports Claude Opus, which works even better than default Claude Code.

...why do we have a bill for $$$$$$ when it was only supposed to be $200 a month/per seat?

Click to expand...

You're talking about an exception, instead of the norm. That was a collection of employees at Amazon who thought it would be fun to see who could use the most tokens. They also had the incentive to keep their jobs, by juicing up the costs. It was not encouraged by Amazon and was ended.

Things are ever evolving, but people need to face the new reality. AI isn't going away - jobs are.

FWIW, if you are a mid or late-career person (most of us on this board), you are currently less likely to be impacted by AI layoffs. It is disproportionally impacting new hires.


The impact is still small, but you can see it rising:

sleepeeg3 said:

I do know how quickly you can burn through tokens. It is not equivalent to one or several employees that it can replace, for actual work.
There are also other AI tools out there that cost a fraction as much. Copilot is $30/month and now supports Claude Opus, which works even better than default Claude Code.

You're talking about an exception, instead of the norm. That was a collection of employees at Amazon who thought it would be fun to see who could use the most tokens. They also had the incentive to keep their jobs, by juicing up the costs. It was not encouraged by Amazon and was ended.

Things are ever evolving, but people need to face the new reality. AI isn't going away - jobs are.

FWIW, if you are a mid or late-career person (most of us on this board), you are currently less likely to be impacted by AI layoffs. It is disproportionally impacting new hires.
View attachment 806836

The impact is still small, but you can see it rising:
View attachment 806837

Click to expand...

You know what also coincides with the end of 2023?

Healthcare, social safety net, infrastructure investment, education. Can't do any of that, all job killers.

Ai Killing jobs in the millions? We can absolutely spend the fuck out of taxpayer's money on that! It's good for everyone with a B as the first letter of their net worth. The people that get paid taxes are destroying the tax base that feeds them

Lakados said:

Without reading the article the answer is yes.

There was a crapload of funding made available to businesses during COVID to hire staff. That funding is gone and a lot of workers just don’t provide any actual value and much of their work can be automated away or simply reassigned to somebody else who AI tools have made more productive.

I know people who exist on both sides of this issue, it sucks, nobody is happy, but it’s the reality of the situation.

Click to expand...

This 100%. There was a lot of over-hiring during COVID. On top of the funding, there was also a tendency to "warehouse talent". There was so much money available that a lot of the big tech companies were hiring people because they didn't want their competitor to. I heard this discussed a lot at the time on financial and tech news media. So a lot of these "jobs" didn't really exist, or weren't necessary. In some cases, they knew it at the time. Now the funding's gone and they can't justify it anymore. AI provides cover for a decision they would have found a way to make anyway.

The downside of this is it's supercharging the Luddite response to AI as some kind of major job killer. It is in some cases, but a lot of firms are reporting they're hiring MORE because AI is creating more opportunities for them.

MrGuvernment said:

I am not saying AI is not a useful tool, it is very useful in the hands of those who know how to use it, but as I noted, excuse, as there are reports now of companies literally going WTF....why do we have a bill for $$$$$$ when it was only supposed to be $200 a month/per seat?

Click to expand...

A lot of that is due to misuse of the tokens by employees, to be fair. There are companies reporting employees are literally using AI tokens to check the weather. Also, there is a trend in the industry called "tokenmaxxing", where employees are basically competing to use the most amount of tokens, but since the objective isn't to be productive with them, they use them for literally everything.

sleepeeg3 said:

I do know how quickly you can burn through tokens. It is not equivalent to one or several employees that it can replace, for actual work.
There are also other AI tools out there that cost a fraction as much. Copilot is $30/month and now supports Claude Opus, which works even better than default Claude Code.

You're talking about an exception, instead of the norm. That was a collection of employees at Amazon who thought it would be fun to see who could use the most tokens. They also had the incentive to keep their jobs, by juicing up the costs. It was not encouraged by Amazon and was ended.

Things are ever evolving, but people need to face the new reality. AI isn't going away - jobs are.

FWIW, if you are a mid or late-career person (most of us on this board), you are currently less likely to be impacted by AI layoffs. It is disproportionally impacting new hires.
View attachment 806836

The impact is still small, but you can see it rising:
View attachment 806837

Click to expand...

This is something that isn't getting discussed enough, in my view. AI replaces a lot of the mundane tasks that are assigned to younger employees, but the negative side of that is those are the tasks that get young people familiar with company processes, build relationships, and become productive employees as their career moves forward. This can effectively cut them off before they have a chance to learn and grow.

HockeyJon said:

The downside of this is it's supercharging the Luddite response to AI as some kind of major job killer. It is in some cases, but a lot of firms are reporting they're hiring MORE because AI is creating more opportunities for them.

Click to expand...

Yes, but it is the AI companies, and the people using AI as a cover, who are to blame for the whole thing. The Luddite response might be a little over the top but at the same time, if the AI companies weren't exaggerating the benefit and other companies weren't hiding behind AI to do their layoffs then the luddites wouldn't have anything to point at.

HockeyJon said:

A lot of that is due to misuse of the tokens by employees, to be fair. There are companies reporting employees are literally using AI tokens to check the weather. Also, there is a trend in the industry called "tokenmaxxing", where employees are basically competing to use the most amount of tokens, but since the objective isn't to be productive with them, they use them for literally everything.

Click to expand...

That is, of course, because some stupid employers set dumb goals for their employees and then are shocked I tell you, shocked! that those employees found a way to game the system.

Ididar said:

Yes, but it is the AI companies, and the people using AI as a cover, who are to blame for the whole thing. The Luddite response might be a little over the top but at the same time, if the AI companies weren't exaggerating the benefit and other companies weren't hiding behind AI to do their layoffs then the luddites wouldn't have anything to point at.

Click to expand...

It's a problem of incentives. AI companies need to convince investors that this is the next industrial revolution in order to raise enough capital. That's easiest to do when you make bold claims like we're on the path to replacing the entire human labour force, rather than saying "hey, check out this cat video I made with AI!". It does seem some AI guys have figured this out and are toning down that rhetoric, but yes, you're correct in that the messaging is definitely not helping the public's perception.

sleepeeg3 said:

Claude Code costs $200/month. A software engineer probably averages $16,000+/month in total compensation. It accelerates what one person can do in many positions by 20-300%+.

Click to expand...

That's a bit inaccurate. For professionals in a job, they're not using the personal subscriptions like you're citing. They're at least on Teams plans, and most likely Enterprise plans (for a lot of the safeguards granted etc.). While those are $20/person/month, you get billed based on token use. My token use at a place where I'm only part time last month is about $1900. I know several others on the team who are near that _per week_. Added up across an org and the full company, it's at least several headcount worth of spend.

My take on the topic/article is that it's a bit of both. And the folks who deeply rely on LLM to fully do their job are basically putting themselves on the "Next to be laid off" lists. LLM is a tool, and can really accelerate a person and a team. It cannot replace thinking, deep architecture, systems design, or nuance about your product or industry yet. It'll likely get there, but, the folks who use LLM to augment their day, giving it the trivial boring work while the human is working on things that are meaningful and impactful to the business are seeing value and faster delivery.
And like others here have said, zero-interest lending, over hiring, and just fat trimming is the name of the game the past ~2.5 years, too. A lot of project and program managers, line managers (engineering manager types), and even M2 levels, are back in the trenches of delivering as well, not "just" managing. Unfortunately that means several things: 1) the quality of management is going downhill, 2) the pipeline for junior level people is really bad right now and no one seems to care about the 10 year picture, and 3) as AI gets refined and better, we'll see more of the same.

The answer is, and always has been, rabid anti-trust enforcement.

sleepeeg3 said:

I do know how quickly you can burn through tokens. It is not equivalent to one or several employees that it can replace, for actual work.
There are also other AI tools out there that cost a fraction as much. Copilot is $30/month and now supports Claude Opus, which works even better than default Claude Code.

You're talking about an exception, instead of the norm. That was a collection of employees at Amazon who thought it would be fun to see who could use the most tokens. They also had the incentive to keep their jobs, by juicing up the costs. It was not encouraged by Amazon and was ended.

Things are ever evolving, but people need to face the new reality. AI isn't going away - jobs are.

FWIW, if you are a mid or late-career person (most of us on this board), you are currently less likely to be impacted by AI layoffs. It is disproportionally impacting new hires.
View attachment 806836

The impact is still small, but you can see it rising:
View attachment 806837

Click to expand...

Does it say where those jobs are? Because outsourcing relatively simple jobs to coding sweatshops in India, Malaysia, the Philippines, etc... was incredibly common, and you had to triple-check that code.
Because it was cheaper to pay $5 US to somebody there to bang out a few hundred lines of slop, and then you only needed to pay the expensive guys in-house for a few hours to fix it.

AI is going to absolutely destroy that industry.

Eulogy said:

That's a bit inaccurate. For professionals in a job, they're not using the personal subscriptions like you're citing. They're at least on Teams plans, and most likely Enterprise plans (for a lot of the safeguards granted etc.). While those are $20/person/month, you get billed based on token use. My token use at a place where I'm only part time last month is about $1900. I know several others on the team who are near that _per week_. Added up across an org and the full company, it's at least several headcount worth of spend.

My take on the topic/article is that it's a bit of both. And the folks who deeply rely on LLM to fully do their job are basically putting themselves on the "Next to be laid off" lists. LLM is a tool, and can really accelerate a person and a team. It cannot replace thinking, deep architecture, systems design, or nuance about your product or industry yet. It'll likely get there, but, the folks who use LLM to augment their day, giving it the trivial boring work while the human is working on things that are meaningful and impactful to the business are seeing value and faster delivery.
And like others here have said, zero-interest lending, over hiring, and just fat trimming is the name of the game the past ~2.5 years, too. A lot of project and program managers, line managers (engineering manager types), and even M2 levels, are back in the trenches of delivering as well, not "just" managing. Unfortunately that means several things: 1) the quality of management is going downhill, 2) the pipeline for junior level people is really bad right now and no one seems to care about the 10 year picture, and 3) as AI gets refined and better, we'll see more of the same.

Click to expand...

Here's the thing people don't realize... they're currently PROBABLY counting the usage by people who don't use it much, or use it properly... which right now, a lot of people aren't actually using their AI tools to their full capacity. Because they don't know how. They don't have the imagination to think of how. In order to actually get the gains that people are thinking AI should give them right now in productivity, though, you often need to use WAY more than that. These subscription services are probably pretty expensive, in order to subsidize the data centers they need to be executed at and their upkeep.

This is probably why a lot of this is going to end up going on prem, not in a subscription.

And I can say that AI can definitely accelerate your workflow. That's just a fact now. I have seen the light after using it more and more.

But what I can also say is that it can actually slow it down. Framing your question is often a lot of trial and error, and a lot of tokens can be wasted on just figuring out how to get it to spit out what you want, in my experience. The Qwen setup I have running locally on my 5090 isn't nearly as well funded as Claude or ChatGPT is, but it has helped me code a LOT of personal projects at this point. But the thing is sometimes to do these projects, I have used up probably millions of tokens PER DAY due to the trial and error nature, and wasted several hours. A commercial LLM will probably hallucinate a lot less than my bandaided shitty local setup, but you have to keep in mind that I'm also thinking very logically most of the time, being a software engineer and an engineering school graduate. I know how to phrase things in a very "this is exactly what I want, these are the exact steps I would like you to take", etc way.

People that have a hard time doing this and properly setting constraints are probably going to have really shitty responses, and they might not even know they're shitty responses, even from cloud based models. And that means they're going to waste a shitton of tokens, or just get wrong responses and waste more tokens later.

StoleMyOwnCar said:

People that have a hard time doing this and properly setting constraints are probably going to have really shitty responses, and they might not even know they're shitty responses, even from cloud based models. And that means they're going to waste a shitton of tokens, or just get wrong responses and waste more tokens later.

Click to expand...

If the companies creating AI wanted to help people they could assist them in figuring this stuff out but, realistically, they make more money letting them fumble around and burn through tokens figuring it out themselves. There is a growing industry of AI consulting companies precisely because people burn through tokens at a ridiculous rate and the executives don't know how to fix it. But, the AI companies themselves want you to burn through tokens because it helps cover the costs of the data centres. Of course, if people get more efficient then they won't need as many data centres ... so catch 22. It is probably a good thing that there is a big push-back on data centres and supply restriction on power, water, and even bodies to build them because that will prevent over-supply of data centres in the long run. If they weren't as restricted we'd likely end up with lots of unused capacity down the road as companies get more efficient at using AI.

May be a dumb/noob question. What about running AI on desktop systems? Let's say your GPU is 8, or mayb 12-16 GB. What kind of models can you run locally?

Does GPU memory determine the number of tokens you can use?

I hope some smart consultant is trying to address those issues.