Management may eventually purge engineers that won’t adopt AI.
Yes and no.
Yes becuse our new C-suites are pushing it heavily and my boss (who only cares that we use AI as much as we need to not draw ire) has warned AI usage is a tracked metric that may be used for perfamnce reviews, bonuses, promotions, and raises.
No because I know my immediate chain of command does not fully believe in AI or are skeptics, and when the bubble bursts the price per token is going to slyrocket. Suddenly minimal AI usage as a tracked metric will probably look pretty good.
Either way once I’ve finished up certain things I intend to look for a new job. I love my team but I fear the writing is on the wall, and even if they try to reverse course I’m skeptical that the company will be able to.
I’m a DevOps engineer (about 15 years) and in a previous life was a software engineer (15 years before that). My employer is pushing hard on AI so I reluctantly started using Claude at times. I must say that I’m fairly impressed when it comes to relatively easy tasks. We’re a large AWS user and have developed a fairly complex in-house set of python tools that encapsulate things like Terraform and Ansible. We have about 15 or so AWS sub-accounts that span logical groupings, so our IAM configuration alone was fairly complex.
I was able to point Claude at our IAM configuration and tell it to create a set of policies/roles to allow a host in one environment to access resources in a read-only manner across all our accounts. Since I’m not an IAM expert it would have taken me a few hours to figure out what it did in under 10 minutes. Two of my team reviewed the proposed changes and were perfectly fine with them.
I’ve also had it write python scripts that do things like call AWS APIs, collect JSON results, and compare it to contents pulled from a git repo of configuration data.
For relatively simple tasks like these it can be a time saver. But you still need to sanity check everything it does. I’ve seen it skip steps (like not applying IAM policies to all our accounts), and when you point it out it will apologize and fix things. But it’s that sort of failure that makes me still be wary of AI. Like why only update a subset of things and fix it only after I point it out? “All” means “all”, not “some”…
For more complex things I’m still very reluctant to trust it. When it comes to that I may use Claude to encapsulate a few API calls, but then I’ll rely on my own expertise to add in all the really important logic.
Same take here. I usually break things down to simpler tasks first and it does better. But it tends to get lost fast if things get too long or too complicated
Well written. This is pretty much exactly how our dev team is using LLMs. Verify everything, but it sure does save time.
I can’t stop management from purging, but management can’t stop me from being one of the few people left whose brain still works and will still possess the skills required to put out their goddamned dumpster fire once the inevitable finally catches up with them.
They’re going to have to beg real hard, though. The kind of begging that has a lot of zeroes before the decimal point.
Be careful. Everyone is replaceable.
I have other skills.
Just a heads up, drawing nsfw horse comics doesn’t count, just learned this the hard way myself
management thinks so
i feel real bad for my coworkers if they try
I honestly can’t believe there are software engineers who think they can just completely ignore this technology and stay relevant.
I’m not an AI bro, I’m not happy with the direction the industry is heading. But saying this technology is useless and refusing to touch it has got to be some kind of coping mechanism. I, for one, intend to adapt and learn how to use the new tools to my advantage.
A very wise man once told me that you will adopt things naturally. Technology that will stick does not need to convince you.
- Smartphones
- Web …
When you need to be “convinced” it is not adequately thought through.
If the hype is real, then there will be nothing to catch up to. All the ”secret techniques” will quickly be irrelevant as the technology is streamlined. Writing CLAUDE or SKILL files manually will be a thing of the past. It will figure it out itself.
Honestly “prompt engineering” is already mostly dead, and in most cases claude already handles its long term memory (claude.md, skills etc…) autonomously. You just have to nudge it here and there to document such and such details you know are important but it’s marginal.
I’ve seen what happens when people use AI to handle anything other than trivial tasks. Calling it useless would be an understatement. It touched code it had no business touching and tacked its own dollar store code onto the end of what it should have been modifying.
My career is already impacted by others using it, whether I use it or not. Those who rely heavily on LLMs produce worse and larger code, and those relying on it heavily are not the best in the first place. It’s turning -1x developers into -10x developers on account of them causing additional cognitive load on everyone else.
As for me? I don’t have FOMO. If I’m right and it’s a bubble that will collapse, then I’ll be better suited to weather it. If I’m wrong and LLMs are all they’re cracked up to be, then I will be able to get up to speed quickly.
If I’m wrong and LLMs are all they’re cracked up to be, then I will be able to get up to speed quickly.
That’s the way I see it too. And is this path occurs hopefully open LLMs will be widely available and at least close in performance to the expensive, privacy invading cloud LLMs.
Although it’s the non-programming related impacts of AI I’m more concerned about.
I’ve been experimenting with qwen2.5-coder:7b and it’s the perfect middle ground. Easily runs on 6 GB VRAM while automating boring stuff and letting me focus on new things.
I’ve been running qwen3 coder 30B and I’ve got to say, I’m unimpressed.
I sent it some lines of python for analysis to tell me what’s wrong with it. it miscounted the index number and said it was fine to run even though it was totally broken.
It constantly trips over itself in mid-sentence stating what it just said was wrong, and then finds a totally different way to be wrong.
it’s about as dumb as a brand new college dropout developer.
I actually really enjoy berating it and calling it stupid when it confidently gets things wrong, so that’s about all I keep it around for.
I think it’s sadly both. Open Ai and Claude will probably die and bring the US economy with it, but the tech is here to stay until the next thing eclipses it.
it will just be like crypto, just small scale for niche stuff. it wont be peddling useless LLM
Not even slightly. Any engineering business going all in on AI is heading down the shitter fast so it’s a useful canary for the competent people on the payroll.
I’m lucky to be in a sector that’s generally too well regulated for slop to filter through, but we’re already seeing fixer upper jobs come in from firms who picked a cheap contract from some AI-first start up and now needs someone to make their new equipment actually work and up to code.
In most cases it’d be faster and cheaper to start over but sunk cost is a bitch.
In most cases it’d be faster and cheaper to start over but sunk cost is a bitch.
- We will need 6 months and $X to rebuild everything from scratch
- But we already have a semi-working thing, can’t you fix that instead?
- Oh right, then it would be 8 months and $2X to fix that.
That’s a generous estimate. When these jobs were a matter of correcting incompetent or lazy human work, you could depend on the problem being a lack of it - lack of compliant methods, lack of error handling, lack of required documentation.
With AI, the volume of work we’d need to comb through has exploded. There’ll be three different subroutines handling the same error, each tripping up the others using it’s own subtly wrong method, and all of it couched in thousands of lines of code handling errors that could not possibly happen.
There’ll be fifty instances referencing an ISO standard and if the standard does exist it’ll have the index wrong or just invent a plausible sounding line to support whatever method was used. Once any error like this is found, ALL of it needs to be verified.
Turning someone’s slop job into something I can sign off on is several times more work than just starting over, for a worse end result. If a client can’t be made to see thats I usually advise to not take the contract.
My company started providing every programmer who wanted it with Github Copilot in like 2023, iirc. I declined it (we were still allowed to decline it back then, sigh). My process of thought was that it takes almost zero skill to “learn” AI tools, but once they’re part of your workflow you become reliant on them and your actual rate of learning stalls. I still wasn’t particularly experienced or good at programming back then (only started my IT career in late 2021) so I wanted to heavily invest in myself and level up my skills as much as I could.
Fast forward to today, and there is a significant skill gap between me and the coworkers I was on par with when Copilot was introduced. I’m still not some kind of superstar programmer, but when it comes to my specific niche (React/Typescript) I’m considered one of the go-to people to consult within my department. Meanwhile some of my colleagues still need almost weekly reminders to use let/const instead of var (yes we have a linter, but they accidentally turn it off sometimes…).About a month ago management started hard-pushing their AI bullshit. Everyone is mandated to install and use Claude Code. So… I did. It took me maybe a day to learn, most of which was spent fiddling with IntelliJ (I took the chance to migrate from Windows IntelliJ to using its Linux build within a WSL, it’s such an improvement!). I did all the mandated Claude tutorials and everything I got out of it is more resentment for my coworkers. This tool really is made for total morons. Even the “advanced features” like writing custom hooks and subagents or connecting to custom MCP servers are just so… stupid. If that’s the most complicated thing they do on a day-to-day basis, I am very much not surprised about their brain atrophying.
Anyways, I am now also the go-to person my colleagues approach when they need help with their Claude setup. Because the guy who self-identifies as competent in AI topics just straight up lies way too much, as he can’t handle being perceived as anything but extremely smart and competent (which just wastes everyone’s time, since finding issues is way harder when he insists his first hunch is always the 100% correct solution and doesn’t admit it when he makes a wrong assumption).
I’m very much not worried about my job security (someone with the capacity to actually understand the code they write is much more valuable and harder to replace, plus I hold a lot of knowledge about our existing product) but I’m certainly having a hard time applying for new jobs. Recruiters and managers really are stupid enough to think that using AI requires any sort of deeper skill, so putting AI on your CV as a skill is basically required nowadays (I still refuse to do it). And the AI bros whose applications I’m competing with are just much more comfortable with lying about their actual coding skill and competence, so on the first scan my resume does end up looking worse than theirs. But I have absolutely crushed every interview that I was actually invited to, because faking technical skill in actual conversation is way harder. I’m not too pessimistic about my prospects of finding a new job, even if the market definitely sucks right now.
I am in charge of policy related to Ai at work, but under different circumstances than most. The people above me asked the question, can we use Ai in a way that adds value to the product? My response was let’s test out a couple of options and make policy based on the results. We gave everyone access to Copilot as we use JetBrains products and I asked every individual what their results were after a month. Four people found that they were able to work faster, the rest said it slowed them down and the hallucinations outright fucked their shit up.
At first I couldn’t get a grasp on how this sanely work for a whole company implemented in a one size fits all manner. Then 2 of the people who used LLMs that were Jr.s left the company. When they had to transfer ownership of the internal projects to other people. I realized the mess inexperience devs with LLMs could do. They left behind untouchable code. That’s when I figured it out, the other two people were Sr.s and you need to know exactly what you are doing to get a net positive from LLM coding.
All three of us were doing slight variations of the same thing, which was very telling of what is going to happen to software over the next year. We would never touch an agent as they just inject flaws and bugs into the code base. We all found Tab complete to be useful enough to leave on, but noted it was worse that worthless quite frequently. What we were all doing was using it as a cracked out search engine through the chat features. Rarely would we accept offered changes directly, but we would manually type sections adapting them to fit better in the code, the way we all used stackoverflow 5 years ago.
I am not saying inexperienced dev can’t use LLMs effectively. You just have to know exactly what you are doing. Now the others come to me talking about MCP servers, agentic coding, whatever new snake oil is being sold to us. Now I am trying to convince these people, who initially rejected LLMs, that they are being gaslit by salesmen into thinking that LLMs are useful.
AI is not the end all be all. Its 1 tool. Its really good at comparinglarged data sets. Astronomers have been using this tech to continually analyze changes in the sky for a long time and its really good at it.
But AI is not inventive or creative. It regurgitates soluyions, and not always correctly
Management really wants to push me into using AI, but I genuinely haven’t found a use for it. It can’t handle complex things, trivial or repetitive things don’t need it, and I have two decades of content that no AI could ever reproduce.
My concern is not that I will suffer from inferior production of code by not using AI. It’s that I will suffer from upper management dictating what tools should be used, without any understanding of the issues or consequences.
No, I have a value driven mindset. If it doesn’t provide value for my work I don’t use it.
My work will speak for itself over the hype. And if it doesn’t? Well someone will
I’m already watching some of my colleagues’ brains turn to mush by just blindly doing whatever the slop machine says. I don’t want to lose my skills in the same way. Years from now, I’ll be ready to retire while the technical debt piles up to critical levels, and demand for skilled and experienced critical thinkers skyrockets; then I’ll have to make the hard decision of leaving the rat race to pursue my own interests, or going back in for one last job for a massive payout.
Or I’m completely wrong and I’ll just be deemed a grumpy old relic by then, and I’ll take my severance and it still won’t be my problem any more.
Severance, you say? AI told HR that you’re not a team player. Non participation = no rewards.
Sorry, I should have been more clear. I meant that I was going to sever one of the HR director’s limbs and take it.
No.
Engineers avoiding its use and solving problems personally are playing the long game. They know that the current LLM tech will collapse - probably due to rising fees and the need to keep growing - and people will be ill-equipped to dig out from under the technical debt – a very real problem.
What makes you think it will collapse?
- AI is not profitable and has no obvious path as of right now ow to become profitable.
- The adoption we’ve seen from so many companies is based solely on the fact that AI is cheaper than salaries. But AI companies are currently running at a huge loss, and the price gauging is inevitable. AI will likely never be as economically viable for the average company in the future as it is right now.
- The future of AI very much depends on it continuing to improve for the next decade at the same rate that it’s improved in the last 5 years. This doesn’t seem super likely given the fact that we got here by training AI on more and more data created by humans, and now that so much publicly accessible content is written by AI, it will be harder and harder to find new training data to improve AI in any meaningful way.
There are just so many big questions out there surrounding AI and it’s masters and biggest cheerleaders don’t have straight answers for them.
Considering the fact that there are open weight models that are pretty close™️ to frontier has me thinking otherwise. Yes I think the frontier companies will probably be face to face with collapse, but capable (and cheap) models already exist and will continue to improve. I think it’s much more likely that companies will simply run their own models (possibly custom agent harness as well) and have all the benefits they were looking for at a fraction of the cost. That being said I do think there will be a significant plateau of capabilities in the next year or so and leadership will realize these are just helpful tools and nothing more.
All that is to say, I don’t agree with your assertion that coders who are not using AI will have any sort of competitive advantage. In fact I think they’re hurting themselves in the long run. I think skeptical engineers who have a foot in both worlds are actually the best equipped for the future. Accelerate your workflow but not at the expense of quality/security.
