Senior devs are quietly using AI for everything and pretending they are not. The hypocrisy is getting old.
Posted by Ambitious-Garbage-73@reddit | ExperiencedDevs | View on Reddit | 48 comments
I have been at this long enough to see the full cycle. When Stack Overflow came out, people said it was cheating. Same with autocomplete. Same with Copilot. And now the same thing with ChatGPT.
But something feels different this time. Senior devs are using AI constantly and loudly criticizing juniors for doing the same thing. I have watched it happen on my own team.
A colleague of mine, probably ten years of experience, gives long speeches about how juniors need to understand the fundamentals and not lean on AI. Then I look at his commit history. He is clearly using AI to write boilerplate, docstrings, test cases. The code has that texture.
I am not judging him for using it. I use it too. What bothers me is the double standard. We are applying completely different rules to ourselves than to people just starting out.
The honest version of the conversation would be something like: here is how I use AI, here is where I still make myself think through the problem without it, and here is why the balance matters. That would actually help juniors build the right habits.
Instead we get vague warnings about fundamentals from people who are delegating the same things they warn about.
Am I just in a weird team or is this a pattern others are seeing too?
Ripolak@reddit
OP's entire comment history is literally posts like this one
DevonLochees@reddit
Starting a day ago, a hard switch from 100% of posts written in a different language to all English posts about AI several of which implicitly contradict each other in terms of what their job/role/experience level is.
Sounds like a sold account doing AI-encouragement spam.
pr0cess1ng@reddit
Some of them even contradicting others lol. Every post about AI should involve mod approval. This type of trash is getting so annoying
gemengelage@reddit
Can we just get a blanket ban against these stupid AI engagement bait posts?
This is one of the few subs where I really value the moderation team, but it seems like they are currently just being completely flooded with this absolute nonsense of AI fluff.
dempa@reddit
> When Stack Overflow came out, people said it was cheating. Same with autocomplete.
Even the post itself smells fishy
endurbro420@reddit
I thought the same. I don’t think anyone thought a website where you can ask programming questions was looked at as “cheating”. That literally makes no sense.
DevonLochees@reddit
Even less sense with referencing autocomplete. Though I guess if this was a junior or perpetual junior I could see them talking about college, where using it to bypass assignments would be cheating.
dempa@reddit
the extra shitty thing is that this reddit conversation (and many others like it) will be fed back into the AI model so it will know how to make the AI engagement post more convincing next time
SamurottX@reddit
OP has nothing better to do all day than talk about AI on the internet
ExperiencedDevs-ModTeam@reddit
Rule 9: No Low Effort Posts, Excessive Venting, or Bragging.
Using this subreddit to crowd source answers to something that isn't really contributing to the spirit of this subreddit is forbidden at moderator's discretion. This includes posts that are mostly focused around venting or bragging; both of these types of posts are difficult to moderate and don't contribute much to the subreddit.
activematrix99@reddit
I have no problems talking to other devs about how/when/where I use AI. Knowing whether or not a senior manager is "AI friendly" is a whole different dance, though. I don't want to reveal too much and suffer blowback or "then why do we even have expensive developers?" pushback.
raddiwallah@reddit
My junior uses AI. I do too. When I ask my junior to back up or tell me what did they change they did achieves, they are blank. I can back and defend all of the changes, even if AI does it. That’s the difference. If you dont learn the fundamentals, you’re going to be useless.
CymruSober@reddit
Well you must be careful here because are you saying that the difference is you can explain AI code on the spot because you understand code or that you check what it does and understand its approach every time?
raddiwallah@reddit
I check and approve each and everything
CymruSober@reddit
Right on! I suspect that the power of a senior with AI assistance is quite upsetting to those without the experience required.
Minimum_Parking5255@reddit
Writing docstringd, boilerplate and test cases are the best tasks to assign a llm
One shorting entire features is not especially if you couldn’t write it yourself, something a more junior developer cannot yet do
Ad3763_Throwaway@reddit
It's no double standard. You clearly do not understand what your coworker is saying.
throwawayacc201711@reddit
Exactly this. Juniors shouldn’t be using AI extensively. You learn pros and cons of approaches and how to design when you actually do it. Juniors need to build those skills. Seniors presumably have that already and can lean more heavily on AI since they can direct it better. A junior using AI heavily is like the blind leading the blind.
Wandering_Oblivious@reddit
The problem is this doesn't align with "move fast, break things" that management imposes. companies don't really believe in investing in employee growth anymore
FlowOfAir@reddit
I agree and it sucks. In those scenarios I would probably say, "okay use AI, but explain the code to me once you push the PR, line by line, and following the execution flow". Unfortunately you need someone to mentor the juniors, something companies don't like to invest into.
dvogel@reddit
I agree it isn't a double standard.
I see the same thing in other industries all the time. Want to be an electrician? Lots of courses about safety. In practice? I see master electricians operating on live circuits often.
Want to be a woodworker? Always use your chisels with a wood mallet... until you've seen every type of wood and a thousand different projects, giving you have a feel for when maybe the handle of another chisel is probably okay.
It isn't whether or not you use a technique. It is how well you can predict whether something will go wrong and recover from those suboptimal outcomes.
drwebb@reddit
I kinda felt the same about Python in the past. Coding using a dynamically typed language for production, with a language that has been known to be hard to refactor (okay modern type systems have made Python a lot better, so maybe my point is a little dated), is actually a job for someone who knows what they're doing. Instead more junior people be like like oh, I just import magic_lib and it just works, so easy to get started, this is the best lang ever.
Farva85@reddit
Bro my 14 year old kid has been talking to me about what subject to major in college. I said “If you want to do computer science, you should start play with AI now.” And her response, “Should I learn how this stuff works first?”
I guess she listens to me working more than I realized, because I say this to my juniors all the time!
Material_Policy6327@reddit
This x100
fanz0@reddit
The problem with StackOverflow, LLMs, etc… is that some people seem to abuse it and not check twice what is being sent for review.
LLMs are very good at fulfilling a goal at all costs so verification is a must. It is the responsibility of the author, not the PR reviewer
thephotoman@reddit
I’m reminded of an old hacker koan:
There’s a difference between a novice using a tool and an experience expert using the same tool.
ryaaan89@reddit
The end of last year I had a project go pretty sideways, I’d been working siloed on it for about a year and turns out my team had been pretty much rubber stamping PRs and then I got a lot of heat for code quality (which IMO were difference of opinion but whatever) and someone senior to me rewrote large chunks of it right before launch. My direct manager, this feels like a bit of a strong word, “accused” me of having AI write it, which I genuinely did not. Fast forward six months and I find out pretty much everyone, including her, is privately using AI for almost everything while still talking about how much it sucks in our public chat. I dunno man, I’m just so freaking exhausted at this point.
Prize_Bass_5061@reddit
What is “code texture”? Could you elaborate with examples.
Is there code that does not have texture?
myusernameisaphrase@reddit
Maybe they can tell from the vibes
/s
FlowOfAir@reddit
It's coarse, rough, irritating, and it gets everywhere
Mundane-Charge-1900@reddit
This sounds like a major cultural problem in your team. Where I work, it is all AI all the time for everybody. Management is raising expectations on output so there’s no time to waste bike shedding over this stuff.
SoftEngineerOfWares@reddit
I can 100% see the difference between my use of AI and a juniors use of AI.
I ask very specific questions to solve small sections of my code and read through the implementation before actually inserting it into my code. Such as by designing the feature myself and have it work on small sections to be plugged in.
I see juniors ask it to write or rewrite a whole feature to solve a simple problem and don’t even think about what has been changed, why, or the outcomes of that change. So when they encounter the next problem they have the code be rewritten again, to the detriment of the previous problem they solved.
pr0cess1ng@reddit
Mods, check op's history
sod1102@reddit
OK, let me see if I can explain this as someone with 4+ decades of development experience. Surely you've heard the phrase "garbage in; garbage out" before? Well, guess what these agentic coding models are trained on? Tons and tons of garbage code that is publicly available. An experienced dev is almost guaranteed to be able to do two things better with AI code gen: 1. provide better prompts so that the code generated not only does the "main thing" you want it to, but also satisfies all of the little NFRs (non functional requirements) that good code should also have (logging, maintainability, DRY, etc.), and 2. an experienced dev should be able to notice shitty code from a mile away, and make the necessary corrections either to the code itself or to the prompt.
rocketonmybarge@reddit
Nobody said it was cheating with stackoverflow. Cause we had to actually code and paste the code and validate it actually worked to solve our problem
Iampoorghini@reddit
I’m a junior and work with bunch of seniors. There’s a big difference in how we use it versus how they use it. I remember my very first PR, the senior dev asked me to explain me the code, and I vaguely did because it was all written by Claude. Since then, I never put up a code i can’t explain.
Generally speaking, the seniors use it as a tool to improve their work, the juniors use it to solve a problem without knowing how to evaluate them, and waste seniors resources to review them.
Therabidmonkey@reddit
I'm literally being tracked with KPIs to track my AI usage. They're pressuring me to use it as much as possible.
merRedditor@reddit
This is the answer. Nobody wants to be labeled as AI-resistant in companies that constantly threaten downsizing while force-feeding AI into everything and telling people that AI-readiness is the only path to job security (which, of course, ends up being a lie anyway).
SorryDidntReddit@reddit
This isn't hypocrisy. Seniors are saying that juniors need to understand the code instead of leaning on AI. This statement is 100% true. You then give several examples of Seniors using AI to accomplish busy work. Seniors don't use AI to architect, prototype, implement, and test code without understanding what's going on. Seniors architect solutions to problems and use AI to execute on their own decisions faster. Juniors don't have the experience to make these larger architectural decisions and they'll never learn to, if they rely on AI to make all of their decisions.
HoneyBadgera@reddit
Both can and should be true at the same time. People getting into the industry absolutely should know the fundamentals, particularly if you’re working somewhere where availability, resiliency, security, throughput, etc are sensitive.
However, the senior absolutely should also be steering those same team members on how to use AI effectively. At the end of the day whether you use AI, write it yourself or illegally outsource your work to someone in another country, you’re still accountable for your code at review time. Finally, the juniors need to also understand the code that’s been written when it comes to incidents and debugging.
grumd@reddit
Seniors who use AI already know the fundamentals. And they're right, you need to learn the fundamentals first. Otherwise you won't be able to 1) know when AI writes shitty code, 2) know when to use AI and when to use your brain (boilerplate, tests, docstrings are all good use-cases for AI).
tdifen@reddit
You're submitting bad PRs that show you don't understand the fundamentals. The senior is using AI but also fixing it when it makes mistakes. You are likely just leaning on the AI and think 'oh it works send it for review!'.
FlowOfAir@reddit
You're in a weird team, and at the same time you don't understand what that means.
Of course, people who know what they're doing can afford to run under different rules than those who don't. Why don't we task juniors with architecture decisions, too?
If a junior engineer is trying to use AI to generate code and is unable to understand what the code is about, I would 100% discourage them to use AI until writing code becomes second nature. If some AI generates code, the engineer needs to be able to understand every single little line that comes out of the fancy autocomplete. If they're unable to figure that out, then they should stick to writing and learning.
NuttyDutchy1@reddit
Maybe you're just experiencing a very specific situation. But otherwise I can see how experienced devs just use it as a tool since they know and understand the shortcomings. While a junior might still be better off doing things more manually to avoid glossing over important details too quickly if AI did the work. They're just people too and might not be bothered to go in depth to improve others
ch4otic-millenial@reddit
Username checks out lol
twisted1919@reddit
Your take is wrong, senior having experience is very different from juniors needing to acquire it.
Alistair401@reddit
this sounds like a personal issue with those devs and not a wider issue that applies to all AI-critical senior devs.
Kaimito1@reddit
Because he knows how to do it without AI most likely and knows when the result is bad so can modify what's needed.
Whereas a junior needs the experience and not 'have it done for him' because they might not know that a test case is overkill or missing