How many of us are on an AI death march?
Posted by ImYoric@reddit | ExperiencedDevs | View on Reddit | 23 comments
If I read the times correctly, a number of companies have entered a new phase in AI adoption.
It started with experiments, with more or less enthusiasm. Now, I witness software companies full of highly-experienced developers that issue internal "you'll get 10x more productive with AI, or else" guidelines, or variants such as "you'll find a way to get the AI to write and review all the code produced by your team, or else".
And if I observe correctly, people on the receiving side of this "or else" are now semi-officially in crunch time, trying to prove that the machine is making them more productive, whether that's true or not.
Is this an observation you share or am I just misinterpreting the times?
Oxi_Ixi@reddit
I talked with many fellow engeneers people on this topic, and my impression is that "reading times correctly" is a virtually impossible. You have too much noise in the media (from AI is doomed to all of us are doomed), you have companies sold to idea of productivity they have no idea how to measure, all sorts of sentiments from love to hate.
I think there are two main dimensions of what goes on in reality. The first dimension is economy. One of the companies I used to work just fired 20% of engineers, and that is not because of AI but because of economics and finance situation. Over last ten years software job market grew too fast, and downfall after COVID was just bouncing back from overhiring and inflated budgets on software silver bullets. As well, economy is not in the great shape itself. So no wonder job market is in bad shape.
Second dimension is our work itself. I think with AI we just entered another round of reassessment of ourselves. How to measure productivity? How should we keep up with technology and how should we adopt it? What is actually the business of the company, as in most cases code just supports it. Etc etc etc. I have seen a few silver bullets, most of those are forgotten by now as such, but they had changed industry too. AI is definitely a big, very big thing, which is changing industry. And things are still settling down, so we don't know shape of the future jobs yet, although we have some ideas about those.
Returning to your opinion, to be honest, I think if a company just pushed engineers for more tokens to use and lays off engineers because AI does job better, it is a bad place to be anyway. There are another companies which are doing fine and still value employees, invest in AI trainings and tools and make sure to get best out of both AI and people. So, I can agree that some companies are in a death march mode. But probably this is a death march for those businesses too. And being laid off from such a ship on the shore now is better thing than drowning with it a year later.
ImYoric@reddit (OP)
This has been a billion dollar question for a while, and nobody has managed to answer it. I seem to recall that Google had published studies on the topic and they concluded that they had no good metric, so they adopted "impact", for better and for worse.
Still better than adopting "number of tokens consumed", as I've seen, which is crazy.
Oxi_Ixi@reddit
There is no right direct software productivity measures for the very simple reason: software is a tool to achieve business targets. Because of that different teams will have different measures. And if you're product or marketing teams don't do their jobs, it won't matter how good and productive software engineers are.
Impact is a tough thing to measure too, as someone on support won't make much visible impact, but that will be the person enabling the rest of the team deploying features for the clients.
And good things to remember: if any performance metric is introduce, people will optimize for this metric, not for actual work goals or quality.
ninetofivedev@reddit
I don’t really feel there is pressure to show increased productivity.
Personally I’m still juggling the same things I always have with software engineering. The only difference is that now the AI writes the code. I’m still doing everything else.
I suspect there is a number of software engineers whose job was to be a code monkey and now they feel like AI does everything for them.
That’s not me. I still have to collaborate across the organization. I have to play politics while different groups argue about the bike shed. I have to campaign for adoption of the tools and platforms that we’re building. And I have to convince executive management to green light the work we want to do for the next half of the year.
ImYoric@reddit (OP)
Heck, I'm a staff engineer, I specialize in fixing things that nobody else can fix, usually across teams, and I regularly push for new projects when I realize that they'll eliminate friction, so I don't think I qualify as code monkey.
Regardless, I do feel the pressure.
rocketbunny77@reddit
I'd say that your take is fairly accurate, yes. The "or else" is a bit more subtle. But ultimately it is "or else"
Kind_Tax4790@reddit
reminds me of when my boss gave "suggestions" that weren't optional at all
Maleficent-Injury-39@reddit
not sure i agree with that point
newyorkerTechie@reddit
Our company started with an increase productivity by 10% using AI or else. Folks had to show off and now they realize we can be a hell of a lot more productive with it. Teams will be warned if they are not “using enough tokens”
ImYoric@reddit (OP)
Oh, yeah, I've seen token usage being used as a proxy for productivity.
That's even dumber than measuring lines of code or PRs/MRs.
newyorkerTechie@reddit
Shit, I’m fine with it. I’ll run four Claude code instances with 4-6 subagents going on each…. I need to try the teams feature again, my last try with it had some issues (rate limits maybe) so I’ve been continuing to manually orchestrating them (makes me feel like I’m doing something).
The thing is, I’m actually getting a lot of shit done. Company thinks so too, they gave me a 10% raise out of the blue. Probably means they are afraid I could get a lot more somewhere else. I’ve been playing and trying yo learn to use these LLMs for coding since 2023 when they were a hell of a lot harder to leverage for actual productivity gain… back then I was basically trying to break even just to learn AI because I figured it would get better.
Fast forward 2026… yeah, if you know how to use AI, you can be hella productive. I’ve been told I do the work of multiple people by many folks, and it’s thanks to AI.
Haha, i sound like a shill for AI. I’ve been encouraging folks to use AI since continue.dev first came out. Before that I was using the fucking chat prompt on ChatGPT
Flashy-Whereas-3234@reddit
If you wanna keep your job, play the game.
Any metric used to measure performance will instantly be gamed.
We were working pretty fast before AI showed up, the main problem of missing or unclear requirements hasn't been improved at all. In fact, now we're compressing research and delivery into the same space, with fewer agile practices.
I quite like AI for finding stuff and doing dumb little fixes, but it's verbose approach and over confidence is a real problem that has to be constrained by better standards, docs, controls, and testing. Shit we were already bad at because we were told to go so fast.
Best of luck to all of you.
experienceddevsb@reddit
This flair is only allowed on wednesday, saturday (UTC). Please repost on an allowed day. Intentionally trying to circumvent this rule will result in a suspension. See: https://www.reddit.com/r/ExperiencedDevs/comments/1rfhdrg/moderation_changes/
AcanthisittaKooky987@reddit
Who cares, do what they want to make them happy, collect your bag, and live your life. They pay you, so do what they say. If you're not willing to do that then simply leave and do something else.
thermitethrowaway@reddit
We're starting the process. Worse yet the people in charge are very definitely subject to the Peter Principle, so they've really drunk the cool aid on this one. One team has been told they must use enough "points" otherwise they aren't doing their job properly so people are generating images etc.
The assurances that the increase in productivity won't lead to job losses were not convincing. Oh and the people who really think this is fantastic are non technical.
Not that I'm against AI, it's just the way they want it to do everything all the time is nuts.
HalveMaen81@reddit
For those not familiar with "The Peter Principle"
https://en.wikipedia.org/wiki/Peter_principle
Tall-Introduction414@reddit
I feel like unions could stop this nonsense, if we had them.
compsciasaur@reddit
Meh, I was on a death march before AI. I actually think it's delaying my eventual layoff at my job. (While also ensuring it?)
vicentezo04@reddit
I personally think AI is a great tool, but it doesn't replace fundamentals.
Too many in the C-suite think LLM coding agents are a magic talisman that 10xs all the KPIs.
Smok3dSalmon@reddit
How do we enable bad managers to fail 10x faster?
PlasmaFarmer@reddit
AI enables managers to be 10x worse in the things they are 1x bad now. One example: measuring efficiency as number of lines of code pushed daily. Or ignoring maintainability for the sake of being fast. They don't want to spend 3 weeks to properly design something because they don't want to waste time but they don't understand that they going to waste months spent on bugfixes and circumventing design decisions in the next years.
Richard-Degenne@reddit
You have the ability of precipitating the fall of these practices by staying the fuck away from any such employer.
SuaveJava@reddit
The main companies driving the U.S. stock market are betting on AI productivity gains, so you will show those gains or be fired. The rest of the economy is in a recession, so devs in those companies are just getting fired while those left behind scramble to use AI to replace their departed coworkers.