I have been reading a lot that 90% of their code is AI generated, companies are pushing developers to use AI as it makes them fast. But I am a little cautious of believing them. Is it true? Also sorry I didn’t find a css career subreddit so I am asking here.
I’m still writing 90% of my code by hand at work. I think if you have total or close to total mastery in your domain, you should probably work faster than AI.
It takes a while for AI to generate code (Opus is pretty slow) and then you have to go review it and do rounds and rounds of fixes. It might be faster to use AI if there were unknowns or if you werent quite sure how to write the code. Otherwise I just find it faster to write it myself.
That being said I do use AI under some soecific circumstances:
- im working in a code base or area of code im unfamiliar with
- Im working in a language in unfamiliar with
- prototyping ideas
- generating boilerplate heavy code
For 1. And 2. I dont usually have ai write code for me. I would just ask it questions like “how do I write X in an idiomatic way in language Y”.
For 3, I have it generate code that I then toss and rewrite if the prototype works.
For 4, this is rare in a good code base. Most of the boiler plate heavy code at work is in unit tests.
I’ve been wary of all those use cases. If I’m unfamiliar with the code base, the language, or have a vague plan then I focus on getting understanding first; AI gives false confidence and I can’t check it’s work if I don’t understand what I’m doing. I’ve never written much boilerplate code as I use templates, written by humans, when starting a new project. The idea of starting from a slop codebase isn’t appealing as you need a good foundation.
AI sucks in my domain. One of our competitors uses it and they say they get amazing results when senior people use the tools but can’t give them to juniors because they keep messing up.
I’m glad it doesn’t work in my niche to be honest. I’m in the frontlines debugging broken code and the last thing I need is bloat. This would actually slow me down a lot. I find it pretty shitty at diagnosing even small pieces of code, and I can’t try stuff like Claude Code because I’m not allowed to transfer some of the code we use to the cloud because it’s under NDAs. But if it can’t get the simple stuff right, I can’t trust it with the keys to our Lambo.
I stay informed about it all to know when/if I should quit software engineering and do something else, but it seems fine so far. It looks like I won’t be able to take it easy in the future and go back to pissing webapp code, which means less opportunities, but oh well.
Oh and I do know two companies that mandate LLM usage. Somebody from my current company left for one of those and hates it.
I am not a professional programmer but it seems to me that the idea that AI is needed to increase the firehose of code being written to “improve” programming and how well computers work is as absurd as the idea that the point of a university degree in a language is to increase the raw amount of words being written in that language.
The point is to convey ideas with language not produce more language, same thing with code, the point is to solve problems not produce ever larger and larger amounts of code with automation.
Something I know without a doubt is that for many people who love language, they desire a great deal less of the fake, hurtful, useless words that drown out the good ones. People who love words and work in crafting and shaping them tend to think it is inherently good to shape useful words not just mindlessly produce combinations of words in as great a volume possible.
To put it in a more abstracted fashion, relying on AI to produce more and more code faster and faster feels like a Jazz musician saying they rely on AI to fill in all the empty spaces they leave between notes with complementary embelleshing notes. The point of a jazz musician is clearly not to produce the most notes possible, it is to convey meaning with notes.
To bring it back to a concrete example, how many times has Google built a new chat program/app from scratch and then abandoned it? Sure there is lots of code there of very high quality, an intimidating amount to be sure, but isn’t the primary job of the programmer here to say “hey, why don’t we stop writing new code from the ground up for every chat app a different part of the company wants and standardize it to a much smaller codebase with a set of customizations different parts of the company can apply to the same core chat program”?
It seems to me a good programmer would be good at framing problems from a perspective that requires as simple implementation in code as possible within reason, not be best at producing the program with the most lines of code fastest that still solves the problem.
deleted by creator
How do you deal with the lack of understanding of the codebase? The company thinks they pay programmers to write code, but in my experience they actually are paying for someone to understand the whole thing. When something goes wrong in production do you just ask AI to identify the problem?
The company thinks they pay programmers to write code
In my company, the C level is pushing hard for AI adoption so everyone is expected to use AI to write code.
When something goes wrong in production do you just ask AI to identify the problem?
Basically, yes. I am no longer programmer but more like software architect or manager. Architecting the design and reviewing AI plan and code.
deleted by creator
Yeap. As a software engineer, this is 100% my experience.
Enterprise Architect here.
This is the answer. All the way.
At my job, employees haven’t written code since the asp classic days and it was garbage back then. This meant almost all new code is written by contractors, which is often garbage. And slow, expensive garbage at that.
Now, AI can at least make better code than the contractors at a fraction of the price.
It also tightens the feedback loop between getting half-assed requirements and getting the deliverables back to those who requested them so they can say how it’s not what they asked for. That process used to take months, now it takes like a day between iterations.
I honestly don’t know where people are working where they say they have tight control of first party deliverables and clear requirements with a cogent SDLC. All companies I’ve worked for have been about 1-2k employees. Are these people working in 10k large organizations where people can afford to be an expert in only one thing and camp on it their whole career?
Also, remember those debugger skills because we’ll all need it.
Large corporate, some AI tools are available, some are optionally tacked onto PR reviews (Copilot on GitHub). No quotas or enforcement yet. Most vocal proponents of AI happen to be the most obnoxious developers or designers.
Writing code was never 100% of the job. The hard part of software engineering is understanding the problem and figuring out the most elegant path to solve it. If AI can do the code-writing part faster, then it’s a good tool to use.
I still spend a third of my week in meetings. I put out on-call fires late at night.
I also spend a good chunk of my time interviewing potential hires. I pretty much expect them to use AI for their code assignments. Including prompt history is a plus if they do. What I do gauge is their ability to explain their code, defend the decisions and know how to adapt to changing circumstances.
I know how to get to this point by starting a couple of decades ago. I do recognise that I don’t have the same grasp of our codebase as if I had written it by hand. I do review everything that gets deployed, but the volume is higher and it doesn’t stick as well.
I don’t know how to get in as a jr today. We’ll know in a few years how it’s done. It’s a new landscape, but if you’re passionate about the field you’ll figure it out.
A lot of the major companies are trying to “embrace” AI. Anyone saying percentages is full of shit. And anyone claiming they got fired for refusing to use AI is also full of shit, they got fired for other reasons, which much have been screaming “I’ll never use AI you fucks!” Haha
Regarding the industry as a whole, AI is a big thing. You need to know how to use it correctly to (hopefully) produce quality code. This is a really rough time in the industry, for many reasons not at all related to AI. The economy is shit, globally. There is a lot of uncertainty, globally. Jobs are absolutely not secure. I’ve been laid off 3 times since 2020. And no, I wasn’t the only layoff. One company got rid of all their IT staff save for the IT director and one (overworked) support technician.
If you’re asking because you want to go into software as a career…………… don’t. Go into healthcare.
I am graduating next month 🥲. A little too late for me to switch careers
Well… good luck. I’ve been doing this for 25 years, and I’ll work for your salary. ;)
You’re lucky it hasn’t happened to you. Yet. I never was a “I’ll never use AI you fucks” type person, more of a “AI written code is not trustworthy by default and should be subject to human review, especially if it touches critical systems, to ensure it’s secure and won’t just blindly wipe out production” type. Even that didn’t agree with their stance.
My former boss literally said to me “AI is so advanced that you do not need to see the code to write it anymore”. That is not true, willfully ignorant, and the equivalent of slamming on the gas on the freeway and closing your eyes. Fuck anyone who believes that.
Well, I don’t have a problem using AI. But I also don’t work for big tech companies with the kind of culture who push it on you. Also, I’m collecting an unemployment check after being laid off because the company can’t afford two developers anymore. So I dunno how lucky I am.
Your former boss is an idiot! Haha
He is! He vibe coded a shitty app, got it on the App Store, and then fancied himself a bona fide developer without having any recent training or experience.
What does his app do? 🤣
It’s since been taken down for who knows what reason, but it was a basic network troubleshooting app. Pings, traceroutes, ASN lookups, that sort of thing. Really simple stuff that a dev with actual experience could probably cobble together in a weekend. He took months prompting Claude.
Healthcare IT is still ok, but seriously DON’T be an actual healthcare worker. Your basically a slave with 5-6 figure college debt
Strong disagree. Getting your BSN is not 6 figures. It also far more stable than tech. There will always be sick people, the world is not getting healthier on average.
Going into software development in 2026 is a gamble. Being a nurse is not.
It’s been creeping slowly into my workplace over the past year. I’ve gotten by without using it myself so far, but there’s been a soft push by management for developers to use AI in their daily work. Experienced devs with a measured approach to AI (“it’s not a silver bullet, but incredibly and increasingly useful”) are given a platform, while AI skeptics are quietly ignored.
Management says they don’t plan on replacing engineers with AI, but it’s hard not to get that impression when a draft for an upcoming company AI meeting has a heading titled “A Bigger Department Without Hiring a Bigger Department”.
AI does not have a foothold in my area, because it still takes far more time to understand what it did, and debug it. Faster to just do it ourselves.
However, there are some programming jobs where people can use what it makes without checking much, or understanding what it does. And while this probably has some cost to pay down the line, if it gets the job done today, then everyone wins.
I know some freelancers, and small shops, that make a lot of money untangling the code made by the above.
Just gonna leave this here.
Not there yet at my company, but management is starting to shove AI down the throats of the more senior engineers at first. I’ve definitely heard of companies where they strictly push for as much AI as possible which is just completely self destructive and delusional.
It sucks that we have to use this crap even if we don’t want it or need it just because the suits see the line go up (even if the line is completely made up of garbage code that will explode one day), but that doesn’t mean you should quit the field. There’s still plenty work to be done, and that will probably go upwards as the symptoms of reckless AI usage start showing up.
The work is worse by all means, you are encouraged or forced to work in a way that strips all enjoyment away, you are forced to nitpick code made by others that you know vibe coded the entire thing, and fixing tons of stupid as hell bugs that a human would probably not make. But still, it takes an actual engineer that knows what they’re doing to be able to clean up that mess and do some actual engineering.
What I fear the most is what comes after the pre-AI senior engineers start leaving or going to retirement, and you’re left with engineers who finished their degrees without ever truly diving into details like one would before AI, starting jobs without learning properly and picking up all the domain knowledge.
Run. Run while you still can.
I recently lost my job because I didn’t want to use AI to write 50-100% of my code as my boss requested.
I technically can’t say I lost my job because of AI as they told me they were laying me off due to restructuring, but a week after I was laid off they were hiring for my exact same skillset with a different title and the caveat that the applicant must use AI to write code. So you do the math.
Dude😭. Where can I even run now? I would be graduating next month
Consider research or academia.
Or on the other end, go into setting up networks & hardware.
Computer science isn’t going away. Computer programming is.
Transition from code to data.
I don’t wanna be choosy, but working to make models smarter does feel like digging my own grave
Not at all what I mean. Working with data has nothing to do with models. Models consume data, but so does everything else.
Right now we are drowning in data. With the AI hype, companies now want to retain their data longer, but optimizing the processing and retreival of said data is an essential art that AI is currently very bad at. Data centers are adding processing power to tgeir infrastructure, but IO is still a massive bottle neck.
i wonder about fresh graduates and how they’re going to survive this job market.
i graduated with an electrical engineering related degree, but it was immediately after the dot com bubble burst so there were no jobs in my field to be had, but i got lucky and found one doing IT and (at the time) there was still strong a enough demand for software development that my IT experience was deemed “good enough” to allow me to enter the field.
22 years later, i got burned out by the culture that software engineers tend to gravitate towards and pivoted back to IT at a non-profit that serves lower & middle income students. i’ve had to work with some of them as part of work-study sort of thing and every single one of them is sharp af – much more than i was at that age and especially so when it comes to ai – but i see every single one of them (justifiably) freak out about their prospects and i feel for them based on my own experiences.
the colleagues at my new firm have been doing this for 30+ years and have never faced layoffs, downsizing, restructuring, etc. and their callous attitude towards fresh grads wrestling this specter is weighing on me just as much as the dominant i’ve-got-mine-fuck-you type of culture that software engineers tend to adopt when in the field.
A friend of mine graduated some years ago from a good uni. He transitioned quickly from actually writing code to just reviewing AI written code. He hates it, and its got to the point he has automated the reviewing process as well. He’s floated the idea of getting into nursing from how much he hates it.
He used to be very passionate about it.
I relate w you though. I graduated from Graphic Design the week midjourney took off. It’s been… Rough.
Dude, I am guilty of writing entire projects with AI. I am not exaggerating, it literally feels like brain rot. Yeah people say now devs have time for high level design but bro, you need experience to know about high level design. Like, I can think of hundred of different ways to design something, but all hundred would be shit compared to what a knowledgeable person would build. Also it won’t be the job most people like
It’s not going to be immediate, but skilled devs are going to be in demand in a few years when the seniors with experience retire and the juniors that never learned to properly code can’t senior.
There are jobs out there, and you’ll probably at the very least have AI tools to use with varying levels of requirements. I have tools but don’t have any expectations to use them. I transitioned from sysadmin to RPA developer to full stack over the past few years with no prior professional dev experience, just one year at Uni and some self learning. So there are spots out there for actual dev graduates.
Here’s the kicker… It’s more about who you know than what you know. Your best bet to get a job is to network and get some sort of referral. Your reference gets your resume read, your resume gets you in the door, and you degree + reference get you a job.
I think what always has plagued software engineering jobs is the office politics. The constant meetings. The ridiculous standards. The monotonous routines. The zig-zagging logistics of upper management. The dead-end dread feeling.
I don’t even work software engineering, but it’s just something I always heard about and kinda seen a bit of.
I write C code for complex embedded products and systems in an industry where change is intentionally a long and drawn out process. Projects take years and support contracts last decades. AI isn’t widely used but it’s available. Most of those that I work with don’t use it at all, and those that do use it use it for cursory tasks like generating reports for management - busy work stuff. There’s talk of it eventually being used for writing and updating tests, the least favorite part of the job. Will it come for more aspects of my job? Probably, and probably sooner than I want.
AI is quite useful. Especially for prototyping and visualizing stuff. Doing array[:-4] or some other nonsense on a column has never been fun for me. It is also amazing at reading manuals and answering questions.
Hallucinations are also becoming more bearable. Accurately understanding the problem and especially the solution is still the most important part of the job.
From a lot of posts here I get that working as a dev in the US is now a total shitshow. But to give you my European perspective (I work in Germany): AI adoption hasn’t been as rapid here. People and companies are more skeptical about it, compared to the US.
I work as a web developer for an employer that is cautious about AI. I can use it, but I am not forced to. Tried it excessively for a few months (agents writing my code, playing a glorified manager and all of that jazz), but I noticed my own skills atrophying and me losing the general grasp of what my code actually does. And even though everyone and their dog claim that the models get better with each new release, I still run into hallucinations way too often. If you are very experienced in a field and you’ve been doing it for over a decade, you notice all of the small inconsistencies and bullshit answers - much quicker than a junior dev who didn’t have that experience yet.
So nowadays I only use Gemini for tool and library research or really simple boilerplate code. For everything else my own brain is the better solution. I am not actively against AI as a technology, but extremely opposed to paying a subscription to some techbro billionaire’s company to keep doing my job. Fuck Altman, Elon, Jensen and the Zuck.
If Ed Zitron is correct in all of his calculations, the frontier models will get so expensive they’ll become unprofitable for a lot of companies, so it would be a stupid decision to rely on them. I am looking forward to one day host good models on my own machine - though that day is not today, when capable GPU’s still cost thousands of dollars.
