This sub is badly out of touch
Posted by brand0con@reddit | ExperiencedDevs | View on Reddit | 103 comments
AI is without question an engineering force multiplier. This is uncontroversial in 2026 and if you believe otherwise, you're closed minded and not trying.
Unsure if the negative AI sentiment in this sub just coming from a small but vocal contingent (ie those who post/comment) or if this era exposes a negative manifestation of "experience" (i.e. unwillingness or inability to adapt and grow) but the posts that rise in this sub seem to be universally overzealous with AI negativity. I'm saying this from all the first hand experience: if you aren't seeing a multiplier in your work, the problem is you. It's probably not even your skills and almost entirely your mindset and attitude.
My team (5, platform eng) ships so much more (and higher quality) software than was ever possible in '22 and prior. It's honestly not even close; there's no comparison. As the staff+ eng of my group, it's plain to see that the tool is just an amplifier of your skill set and ingenuity from the before times. Those who don't bother to put in the work to understand a system's architecture, implementation trade-offs, or lack the wherewithal to call out stupid mistakes or paths taken by LLMs will struggle. Those same people struggled before in exactly the same ways, just slower.
By contrast, those who are able to approach problems and tools with an open mind, think critically, and look for opportunity will find them in abundance in this era and beyond. The craft has changed. Mourn it if you need to do that. But it's wishful thinking (at best) to expect practices to revert to some former way. You're lying to yourself and doing others a disservice if you're broadcasting hopium.
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.
cellularcone@reddit
Is that why every app and website is shit now?
brand0con@reddit (OP)
Every app is all of a sudden shit because AI. Tell me you want a scapegoat without telling me you want a scapegoat. This is religion.
TheBoringDev@reddit
Irony so thick you could cut it with a knife.
brand0con@reddit (OP)
The difference is evidence. Time will be a harsh mistress to your denial.
apartment-seeker@reddit
The degradation def started before AI
Regal_Kiwi@reddit
Yeah it just accelerates the decrepitude.
cellularcone@reddit
True but it’s accelerating.
BobFellatio@reddit
No, they were shit before as well
Sheldor5@reddit
Microslop ftw
jakeStacktrace@reddit
If I don't agree with what you say I'm close minded and not trying.
I'm open minded enough to think it might be a skill issue in both our cases which is something you took off the table since you are close minded, over generalizing and offering nothing but a false dichotomy.
brand0con@reddit (OP)
Try to follow along. My argument is, if you disagree with:
Only then:
Disagreeing with this can absolutely be attributes to a skill issue which imo clearly rolls up into the not trying half.
jakeStacktrace@reddit
If I didn't know coding vibe coding would be a force multiplier. So it is for some. But if you can already write that code it's less useful. I find I have to rewrite a lot of the slop it makes so if is less useful but I wouldn't know better if I wasn't as good as I am.
The skill issue on the other side would be me not being good at applying ai assistants, making rules files, or ignorance of the tools available.
Neither of these skill issue cases has to do with not trying or being close minded. It's OK I would have trouble finding edge cases like that in my logic too if I haven't been doing this for so long.
brand0con@reddit (OP)
Dunning Kruger is one hell of a drug.
jakeStacktrace@reddit
I use AI to code for me at work all the time, I just get tired of others unable to use logic.
There is certainly one thing that beats my programming skills every time and that's the bullshit others believe.
You sound like you never even took formal logic honestly.
brand0con@reddit (OP)
That youre talking about college courses, and hyper focused on your superior programming skill and logic says everything about where you are in your career. We are operating at very different levels.
jakeStacktrace@reddit
Embracing logical fallacies doesn't work even if you are commanding AIs. And you still have to know what you are doing. At some point my skills may not be useful but we are not there yet.
Inevitable_Zebra_0@reddit
I don't know... I'm using agents for coding, sure, but many times it does such stupid stuff, like, inventing a wheel instead of using an obvious library for the language to complete a task, or violating SOLID, DRY and OOP principles so blatantly that a lot of times I realize I could've implemented a feature faster myself than just relying on the coding agent and then having to interpret and try to fix what it did. And it's also taking a lot of time to run, e.g. it's not rare for the process to run for 10 minutes, before I look at what it did and just have to revert it all back because the agent took a completely wrong turn (such as one of the examples above). And that's me using the latest models, btw. Several areas where I found them to be very useful at are explaining legacy codebases and writing boilerplate routine code like unit tests, configuration files, python/SQL ad-hoc scripts.
Historical-Essay-128@reddit
No one is denying AI is a great tool in a developer's toolbox. But "AI made me x11 engineer" is usually just a hyperbole.
TheBoringDev@reddit
To be fair I’m denying that. It’s a mid tool that’s at best situationally useful.
boringfantasy@reddit
We didn't need to code faster. We actually probably should've slowed down. Agentic coding makes almost everything worse. LLMs would've been best placed as judicious reviewers that sit alongside you.
brand0con@reddit (OP)
Was writing software never a bottleneck in the process? I'm genuinely interested in understanding where you think the bottleneck was in your business.
yxhuvud@reddit
The bottleneck is, and has always been, understanding the problem you are trying to solve and how it relates to existing solutions.
If_I_Could_Just@reddit
Extremes are almost always a red flag. It’s not useless. It’s not so amazing that you don’t have to watch it. It’s a great tool for a human.
airemy_lin@reddit
You’re right but how many people are out here spouting the same stuff from two years ago as if AI hasn’t advanced since that time?
We were all right on blockchain and NFTs being a dumb fad for grifters and now some people are unable to see that LLMs are NOT in the same boat and that an insane amount of capital is flowing to it.
yxhuvud@reddit
The AI bubble burst will mean it will go away as much as the dotcom burst made the web go away.
If_I_Could_Just@reddit
Be happy that you’re competing with people who are less desirable to employers. They don’t want to work with people who can’t squeeze the positive from the new tools.
brand0con@reddit (OP)
Exactly! Anyone who claims they let agents burn tokens for hours on problems is outright lying. Anyone who insists this is a fad is just as dumb just on a different vector.
MercyEndures@reddit
I do multi hour Claude planning sessions to come up with detailed, like 20+ page design docs, then let an agent orchestrator yolo on it for 5-6 hours with a dozen or more agents and then have a working first draft of a new service.
If_I_Could_Just@reddit
I will say, as I improve my guardrails, the longer I can let it run. But eventually, errors and compound and it goes off the rails.
bold_snowflake@reddit
Strong emphasis on tool. The companies that succeed long term will be the ones that understand it's a capability enhancer for skilled engineers. The companies that fail will be the ones thinking it's a replacement for expertise.
norse95@reddit
All this sub tells me is that nobody knows what the fuck is going on
tuna_safe_dolphin@reddit
This is actually true. No one knows what is going on now nor do they know what will happen over the next 1-10 years.
SolidDeveloper@reddit
You couldn’t just say “hope” here?
dual__88@reddit
"My team (5, platform eng) ships so much more (and higher quality) software than was ever possible in '22 and prior."-oh, so your salaries increased 10 fold then?or at least the company's profit?
brand0con@reddit (OP)
Wait, help me see where I said our salaries all 10x'd. Was that my claim or the part you fabricated? Do you really think effort or output has a linear relationship to comp? Sweet summer child...
R3mainz@reddit
If you couldn’t tell he was being sarcastic I’m surprised how you became a staff+ engineer of a large org, that’s pretty simple context use right there. Also if you don’t think delivery of products with value isn’t directly tied to comp, I have a bridge to sell you.
brand0con@reddit (OP)
How could I have possibly missed the sarcasm jumping off the page? Serious question (to an unserious person so my expectations are low).
It turns out reading tea leaves on reddit has basically zero correlation to leveling. Shocking, right?
Snoo34567@reddit
Are you being sarcastic? Tertiary sarcasm always throws me off.
madethisforcrypto@reddit
His team is still just as vulnerable to get shaved down in layoffs .. he loves making higher ups happy with his AI glory stories.
bold_snowflake@reddit
I don't t know many engineers who claim it's terrible. The trouble I see is the magnitude of which it's a force multiplier. People thinking it's a 10x are frankly deluded right now.
You claim your team is shipping more and at higher quality. Can you elaborate? - how are you measuring quality? - what is your definition of shipping more? More code? More features? - how Greenfield/brownfield is the product? - how large/complex is the system?
It's not all apples to apples comparison.
Repulsive-Hurry8172@reddit
For me it should be ultimately, is business making money off their workers using the tool.
A lot of us feel 10x when those tasks has no business impact.
If_I_Could_Just@reddit
It can be 10x at specific tasks. But it doesn’t mean the whole job is 10x.
weakestfish@reddit
The problem my friend is that it’s a force multiplier in both ways. Good engineers can make more good stuff, bad ones can make a whole lot more bad stuff.
And the unfortunate truth is that most engineers are mediocre or bad.
Speaking for myself but that’s where my negative sentiment comes from. I’d rather sacrifice the gains I get from it (which are real and useful) if it meant I didn’t have to deal with the absolute dog shit some of my peers create.
Snoo34567@reddit
This is a less nuanced approach masquerading as interjecting nuanced to the OP’s criticism.
Without getting in-depth. You are stating you are use to ignore the inefficiencies of your team, but AI has made it so you now have to address them. The average SWE not being “good” at their job has nothing to do with AI. Unskilled people causing more damage than they are worth when given access to new tools is a people that has occurred and been solved over and over again. The answer has never been to blame the tool.
The OP did not over look your observation. It was not relevant to the overall point with how easily your issue can be rectified in the future.
weakestfish@reddit
What? I don’t think you understand and frankly the first sentence of your comment makes no sense.
My point is that it is inevitable that we end up with shitty peers. That’s an unavoidable fact of life. This particular technology allows them to produce a huge output that gets applause from managers because they just see velocity go brrrr. Meanwhile, thoughtful and careful engineers are getting 1.5-2x boost, but are still taking their time and being detailed. This does not get applause from management and in fact we get asked why we aren’t going 10x like Jimmy Vibes in accounting.
My final point is that even if AI gives me 1.5x-2x speed, I’d give it up to stop the slop above.
Snoo34567@reddit
The solution lies in developing more comprehensive training on AI usage, implementing standardized checks for edge cases, and establishing guardrails to minimize AI Slop. Your point is a common issue that arises when introducing any new tool, work structure, or system. We are currently at the stage of implementation where you would anticipate an increase in AI Slop. Over the next 2-3 years, we can expect to reduce the AI Slop caused by subpar SWE to leave only a 1.5x-2x (or greater) boost.
This is essentially what my initial sentence was implying. Your comment addresses an issue that was already known to occur over 4 years ago, and plans have already been made to rectify it outside of your purview. The notion that poor SWE can lead to more Slop is the natural growing pains of implementing AI software, and it will eventually resolve itself within the next 2-3 years.
At the end of the day, you’re not wrong. This is a management and organizational structure issue. Management doesn’t comprehend your job, and your coworkers are less skilled and knowledgeable than you’d like. Therefore, you don’t want to provide them with the tools that would make your life more challenging. That’s completely understandable. It’s one of the primary reasons why progress is difficult.
I see your point. I sometimes tend to be overly optimistic and believe that people can only be dumb so long, despite my own lived experiences.
therealslimshady1234@reddit
Well said ngl
weakestfish@reddit
Thanks mate
Sheldor5@reddit
AI is like a fart, if you have to force it, it's probably shit
therealslimshady1234@reddit
It’s the only tool in history that has been forced upon us by management
Snoo34567@reddit
This is the most untrue statement, I’ve heard all year.
Deaths_Intern@reddit
Weird take man, definitely not true
therealslimshady1234@reddit
How is it not true
Deaths_Intern@reddit
Jira, confluence, Jenkins, GitHub actions, linters, just to name a few... I can just think of so many counter examples over the years where management pushed the trendy "new" thing and derived metrics from it.
It's not a new thing; they hear about something they didn't hear of before that their peers are talking about, think it will solve all their problems, and push it hard for their teams/orgs.
Same shit, different day
therealslimshady1234@reddit
Oh come on now, you cant possibly compare an issue tracker to AI. Thats so disingenuous even for an AI bro
Deaths_Intern@reddit
I'm not comparing an issue tracker to AI man, I'm saying that this is far from the first time I've seen management push for the use of a tool.
therealslimshady1234@reddit
I was referring to having some failed-upwards senior manager tell me how I should actually program.
Or even worse, non-engineers. I now have designers telling me I should use (their) AI generated frontend libraries. This industry has become pure cancer
airemy_lin@reddit
I mean this doesn't feel all that different from agile/scrum being pushed down everyone's throats in the past. But the difference is I actually like using AI tools.
Deaths_Intern@reddit
Yeah that would be annoying, fortunately I don't work somewhere like that
Maybe do some interviews, good luck out there
brand0con@reddit (OP)
If management is forcing it on you before you've given it a shake, that's your fault, friend.
I'm as annoyed as any by execs giving proclamations about how we should "revolutionize how we work." Many of us have been doing just that throughout our entire careers.
naked_number_one@reddit
DORA report the other year said it’s indeed a multiplier - good teams become better, bad teams become worse
therealslimshady1234@reddit
I still havent been able to find any team that has become better.
If you look at the Claude Code source leak from a few weeks ago, you can see that even that was pure slop. And they are a bunch of "elite" 500K-1000K TC engineers with unlimited access to the best unreleased models in the world
No, it seems that AI just regresses you to the mean and adds a small velocity bonus, which eventually gets negated by the inevitable tech debt that creates
naked_number_one@reddit
Well the dora is a reputable research institution. What they see is that teams with good processes are able to amplify performance. If you don’t see this around only means those teams have low vaseline.
Take into account they they use objective measures like delivery velocity, release stability, etc. rather then subjective “I don’t like this code”
All their research are public. You can see their consolidation ans well as metodology
therealslimshady1234@reddit
Although DORA was able to find a 33% speed up in tasks completed, they were also seeing a 54% increase in bugs, so I don't think your study is really all that helpful to the AI cause. This simply confirms my suspicion that AI has made trading quality for speed trendy.
naked_number_one@reddit
This is the whole narrative of the document. The amplification lies not in typing faster, but somewhere else.
Interesting-Monk9712@reddit
Logical_Newspaper_52@reddit
IMO it’s quite opposite: the better and the more experienced a dev is the more benefits from AI they get. they know what needs to be done and what’s a waste of time. good and bad devs had to spend significant amount of time for routines, so their output was averaged a bit. now AI can handle a lot of routine
Interesting-Monk9712@reddit
If you problem is writing code, you are not a great dev.
Logical_Newspaper_52@reddit
try to reread my comment a couple of times
Interesting-Monk9712@reddit
Try using AI, maybe you get it then.
Logical_Newspaper_52@reddit
it says your English is rubbish
Interesting-Monk9712@reddit
If you have repetitive tasks or so called "routines" you should have already automated if you are worth anything or developed something, if just waited for AI to come, sorry mate, you are not a great dev.
Knoch@reddit
I feel like all the good devs I know get massive benefit out of it. More than mediocre devs.
Interesting-Monk9712@reddit
AI can help you write code, but if writing code is a problem for you, you are not a good dev.
The real issues are anything but writing code.
therealslimshady1234@reddit
No bad engineer has even become good due to AI
Interesting-Monk9712@reddit
Not good, but better.
therealslimshady1234@reddit
Which makes them a bigger liability than before
Interesting-Monk9712@reddit
Not if your company is fine with it, most companies have mediocre devs, making mediocre products.
Even for FAANG you can see how bad their products have become.
therealslimshady1234@reddit
Yes, everything is now shittier. The weird thing is, we dont even seem to get new features at a faster pace. It seemed to have remained exactly the same. So what are all those hundreds of billions of USD doing?
brand0con@reddit (OP)
100% - AI absolutely enabled inexperienced and sloppy devs ship more software in the wrong directions. That's not making them good. Creating the right thing doesn't become obvious because AI
boring_pants@reddit
I don't think you know what "uncontroversial" means. Perhaps you should ask chatgpt about it.
In fact, why don't you delete this post and just have a conversation with your LLM of choice? It'll give you all the fawning responses you need and tell you how right you are, and you don't have to deal with all these tedious human beings who, by disagreeing with you prove how close-minded they are.
Chatgpt would never do that!
So go and talk to it.
madethisforcrypto@reddit
The most we can do is have an eye on who pushes out bad code. If your whole team pushes out bad code the manager is f*cked.
If your seniors have a brain, then people pushing out bad code should be punished more than before.
MercyEndures@reddit
Do you ever meet engineers who think AI can’t code offline?
I do meet non-engineers who have absorbed narratives, and then I can walk them through what I did yesterday and how long it’s been since I actually authored a line of code.
brand0con@reddit (OP)
This is the true answer and what I suspected. Anyone who makes contact with AI genuinely in their day to day SWE practice understands. It's only the terminally online who bring untold baggage.
SrDevMX@reddit
Using AI requires nuance because these powerful models lack contact with the messy reality out there. To use AI effectively, you must balance its productivity benefits with careful human oversight, you can easily introduce logical bugs, security holes that hide in plain sight but are harder and more expensive to find.
mq2thez@reddit
16 YOE, across a mix of platform, web product, and dev productivity work.
I was strongly anti-AI for a long time, but Opus 4.6 really changed the game. The quality got good enough that it really helped me work faster and better, so now I use it.
That said, it’s far better for platform work (which is often large-scale mechanical changes, things which can be deterministically examined, etc) than for product work. For building new product, I find that it’s mixed at best, especially when needing to thoroughly plan complex changes. Even worse, it’s frequently lazy or dishonest and hides it.
The real problem with AI is what management expects with it. I was thrown onto a critical project with massive scope and very short deadlines as the solo UI engineer. The deadlines require me to, essentially, merge 2k LOC per day for several months. We’re a little over 2 months into that process. Requests to cut scope were ignored until I was yelling and jumping up and down, and because I had zero context on anything I was doing when I was assigned to the project, we were deep into thing before I realized how truly fucked things are. With a week before a critical deadline, I was pretty behind and the project leads were finally asking if they need to get me more help, but at that point, it would have hurt the deadlines more for me to have to onboard someone else. All of this is treated as fine, and normal, in the times of AI. And because I’m a pretty talented engineer using AI as a force multiplier, I’m mostly keeping the wheels on.
But when the project ships, I’ll be leaving behind a massive pile of AI generated code I do not understand, and can’t help someone else maintain. I’ve had AI writing docs and tests as I go, but there’s no time to really review those or be sure if they’re helpful. I can’t imagine anyone else adding to this system without even more AI. It’s a huge disaster in the making, but no one cares because we have to hit the deadline. A 6-9 month project from before AI for a team, done by one person in 2 months.
adappergentlefolk@reddit
what are the new unmissable ai written software products
was all good software already written before ai?
Adept_Carpet@reddit
It feels like it when I use it, but I'm looking around and I'm wondering "where's the new stuff?"
I see a swarm of new prototypes that are cool but not usable products yet. The mature products I use, I'd say it's about 50/50.
Some of them have begun to iterate faster or have implemented long awaited features that were clearly being avoided due to cost/difficulty, but some have gone the other direction and have become flakier/less useful.
So it's clearly 10x or even 100x from idea to prototype, but it seems like the systems that do real work are moving about the same speed they always have.
tomkatt@reddit
I take anything said by someone with a hidden profile with a bucket of salt.
Not saying AI isn’t useful, though I have concerns about the costs and long term outlook for skills and junior development.
airemy_lin@reddit
Companies were already not hiring juniors before AI tooling really took off anyway.
Death of ZIRP has probably choked out the junior pipeline for the near and medium term future until we see a shortage of mid level and above talent soon.
w-lfpup@reddit
I love ai yum yum yum I say with a suspiciously "getting paid for using AI"-sized bulge in my tummy mmMMMmm yummy
therealslimshady1234@reddit
All tech boards are flooded by Anthropic's sockpuppets nowadays
log_alpha@reddit
The sentiments are negative because AI only helps billionaires become more rich, not the average worker. More money for them, joblessness for the rest.
Like look at the tech industry. If 5 people are doing the work of 10, who really benefits from it?
apartment-seeker@reddit
But this is the problem with modern American capitalism, not AI
Key-Organization3158@reddit
We'll become more of a proper engineering discipline. Civil engineers don't build the building. Mechanical engineers rarely touch a lathe. And now software engineers won't directly write code.
There was a time before compilers, packages, garbage collectors, and runtimes. Each one carved off some grunt work and ended up doing it better than we ever could. This is no different.
Agent7619@reddit
Regardless of how great or not great AI is from a productivity perspective, from a physics and economics perspective AI in its current form is unsustainable.
kayakyakr@reddit
I've noticed companies are suddenly getting very cost conscious. Just an interesting shift
Petur06@reddit
Well, to be fair, 0.X is also a multiplier.
Rascal2pt0@reddit
It’s a multiplier for sure…. But we’ll have to see what happens as people are pushed to lesser models and have to ration because of cost.
-no_aura-@reddit
I don’t know anyone who thinks it’s useless. The tools are great, I use the every day as does my entire team. The most pushback I see here and IRL is against non technical C suite/MBA types pushing it hard as a silver bullet to be able to slash their dev team numbers and budgets.
QuitTypical3210@reddit
Wats ur product
Jobidanbama@reddit
Good one