Forced to vibe code at work, programmers say their skills are deteriorating

[Daniel Sims]

Cutting corners: Tech giants constantly boast about how programming, among other tasks, has seen massive productivity gains from generative AI. However, recent complaints from programmers are beginning to echo prior studies highlighting the detrimental effects of vibe coding, such as increased debugging time and cognitive decay. Software engineers remain divided over whether a middle ground can be found.

Coders from various companies recently told 404 Media that their initial curiosity about vibe coding has soured as they feel their skills deteriorating while technical debt mounts. Many developers who aren't being forced to use AI are returning to coding by hand.

The programmers who spoke to the outlet said they are either under explicit AI mandates or that AI usage factors into their performance reviews.

Similar stipulations have led Amazon employees to inflate their AI usage. The comments echo recent boasts from companies such as Microsoft, Spotify, and Anthropic about how much of their code is now AI generated. Last month, Google claimed that AI now writes 75% of its code, while the figure has reached 90% for Anthropic.

While tech giants argue that AI allows each programmer to do more, coders' complaints primarily highlight two issues: first, generative AI produces too much code for humans to debug efficiently. Some developers either grow more exhausted from analyzing AI-generated code or cannot understand it at all, and ultimately ship products without knowing whether they are flawed.

The second problem is that programmers can feel themselves becoming worse at their craft. One respondent compared the feeling to when people stopped memorizing phone numbers after they began using cell phones.

Commentators on Reddit said that even if coding by hand takes longer, it gives them a deeper understanding of the code, enabling them to debug it more effectively. Some programmers worry that, while experienced coders will likely adapt in the post-AI era, vibe coding might endanger the next generation's ability to analyze code or even perform basic tasks.

Meanwhile, others argue over whether careful programmers can strike the best balance between the pros and cons of vibe coding. A recent blog post by a developer who pledged to abandon the practice claimed that AI can effectively write code for individual features but cannot handle architecture and loses context when projects grow beyond a certain size.

Although the programmer has stopped using Claude and returned to Rust, he theorizes that AI tools remain usable if developer prompts include clear boundaries. Some commentators on Hacker News agree, while others remain staunchly against vibe coding.

Permalink to story:

Forced to vibe code at work, programmers say their skills are deteriorating

 

[Axeia]

This has long been a problem trough a disconnect between upper management and the actual developers for ages.

Productivity is often measured simply in lines of codes written, maintenance or debugging be damned. AI is very good at spitting out big blobs of code making you look super productive... until things go wrong. If you don't know how all the generated code works you have a problem.
Understanding how these big blobs of code actually work imo you only do from having done it manually before. And yeah I agree, that's something that deteriorates once you stop doing it.
I wouldn't be surprised if in 10-15 years we're stuck with a few people that refused to use AI maintaining AI generated code to keep the world running and being paid extremely well.
Just like how people fluent in COBOL can make tons of money because the world relies on it but there's not many that can write it.

 

[p51d007]

Duh!
How many articles have you read where kids language, social, intelligence, cognitive levels have dropped
because they are "doom scrolling" all the time. Instead of using their brain to try to sort through
something, they just "ask google".
It's like anything else...your body wears down faster if you don't keep it in shape.
Same thing can probably happen to your brain.

 

[RudyBob]

Duh!
How many articles have you read where kids language, social, intelligence, cognitive levels have dropped
because they are "doom scrolling" all the time. Instead of using their brain to try to sort through
something, they just "ask google".
It's like anything else...your body wears down faster if you don't keep it in shape.
Same thing can probably happen to your brain.

My thought was if they can feel their skills slipping then it's time to do some coding on their own

 

[TooQik]

This is no different to having to debug another human's code - if you didn't write it, you need to slowly move through it to understand what it does exactly. This is why comments and structure in code are important.

It's bad enough trying to find your own mistakes in code, let alone trying to find mistakes made by someone/something else.

 

[toooooot]

I am naturally bad at math. When I was in school, I learned just enough to pass classes.
But since I have not use any of those things they taught us, I can barely recall anything.

I mean it is only logical that the people who stop coding forget how to do it without AI typing it.
But I think the worse is still ahead, when they do not remember how to AND also forget
what it is that AI "vibed" and why it is not working.

 

[daffy duck]

I guess I would be black-banned form the industry as I would tell them all to shove their AI up their deluded a$$. Not on a cold-day in hell would I let AI take over my coding.

 

[Theinsanegamer]

My thought was if they can feel their skills slipping then it's time to do some coding on their own

And when are they supposed to do this? They already spend all day coding and have to use AI tools or get fired. Are they supposed to sacrifice what little off time they get doing even more work?

 

[abactuon]

So, how do things work now when we need to create some unique program, for example, for the Artemis mission?
Will nuclear reactors, airplanes, and the like also be controlled by programs created by AI?
What a fantastic thing they've done. Gödel's theorems have stopped working.

 

[MarkHughes]

Duh!
How many articles have you read where kids language, social, intelligence, cognitive levels have dropped
because they are "doom scrolling" all the time. Instead of using their brain to try to sort through
something, they just "ask google".
It's like anything else...your body wears down faster if you don't keep it in shape.
Same thing can probably happen to your brain.

As I have gotten older I have come to realise the term, Use it or lose it, describes a very real problem.

If you stop running you eventually lose the ability to run, If you stop thinking the same thing happens, I have often described it as your whole body is geared towards existing in the lowest energy state that it can, So if you leave the room it will turn out the lights so to speak.

And as a developer I have noticed some laziness creeping in since using AI to help. I am trying to find a good balance between it helping me in a genuine way and it taking over my thinking in a harmful way.

 

[user556]

... I am trying to find a good balance between it helping me in a genuine way and it taking over my thinking in a harmful way.

Use the LLM as another search engine for providing canned examples. Then write your own version for the program you're writing.

 

[MarkHughes]

Use the LLM as another search engine for providing canned examples. Then write your own version for the program you're writing.

We do that, and I have it check the code and give opinions, And it's not bad at unit tests either. All our code is also reviewed by people and the results are tested with an automated system... We aren't allowed any vibe code except for prototypes and internal tools.

 

[GeoffreyA]

It applies to most skills, not only programming. The more we rely on something outside, appealing to indolence, we grow rusty and lose the ability to do it well. We were given bodies and brains to use.

 

[LemmingO]

Funny how we're allowing AI to do the coding, and then spending a massive amount of time debugging the **** out of the slop.

 

[Endymio]

Funny how we're allowing AI to do the coding, and then spending a massive amount of time debugging the **** out of the slop.

Wait, I thought everyone was being fired due to AI? Who's spending time debugging?

It's a wonderful example of Orwellian DoubleThink that the anti-AI crowd simultaneously embraces the two contradictory ideas that "AI don't work!" and "AI is taking all our job!".

 

[wiyosaya]

The decision makers who are forcing the programmers to use AI should, themselves, be forced to use AI in order to get a taste of the ramifications of their decisions. As I see it, the decision makers are jumping on the AI fad without understanding the consequences of their decisions. IMO, not even remotely a good move.

 

[wiyosaya]

Wait, I thought everyone was being fired due to AI? Who's spending time debugging?

As I see it, that's not the issue here. The issue is that the code that AI is producing is garbage as expected. And the decision makers are jumping on the AI bandwagon without any understanding of the fact that the AI produced code is complete crap.

Thanks for yet another strawman argument.

 

[RudyBob]

And when are they supposed to do this? They already spend all day coding and have to use AI tools or get fired. Are they supposed to sacrifice what little off time they get doing even more work?

It's not work if they like what they are doing. They care enough about their skills to bemoan the loss but not enough to insure that they don't lose those skills. I understand your point and I agree that if someone spends the day doing something they don't want to go home and do more of it just don't complain about it

 

[Tech420]

This happened to me with navigation when GPS was invented. I used to print paper maps off mapquest if I wanted to go somewhere, but now with GPS I don't have to think about navigation. I don't even know which way is north/south/east/west in a city anymore. Same thing will probably happen with AI coding. You'll still get results, and better and easier than before, but you won't k now how to do the coding yourself, and nor should you. Coding isn't a natural human ability - there's no reason you should have to talk to a computer with a programming language - you should be able to use plain language.

 

[Axeia]

This is no different to having to debug another human's code - if you didn't write it, you need to slowly move through it to understand what it does exactly. This is why comments and structure in code are important.

Just kidding.

AI is pretty trash at writing useful comments unless carefully instructed on how to do it though. It likes doing things like

// Fetches index page
$this->fetchPage('index')

Or foregoing them entirely.

The more code I write and read the more I come to realize that comments the majority of the time are rather useless. Good code should be self documenting most of the time, by the time you're writing a comment you should ask yourself the question whether you shouldn't simply be splitting up functionality in more bitesize pieces and/or using more descriptive names.
Also greatly reduces the chances of things going 'out of sync' as it's much more intuitive to change the name of a method when it's functionality changes than it is to re-read comments above or sprinkled throughout it and update them.

Where AI can however shine is in IDE friendly comments in weakly typed languages. If you expect a certain type of variable or return value and it can be specified in a docblock then by all means let AI do it for you.
Thankfully PHP nowadays can handle most of it through the syntax rather than having to rely on comments.

 

[Rumiku]

Wait, I thought everyone was being fired due to AI? Who's spending time debugging?

It's a wonderful example of Orwellian DoubleThink that the anti-AI crowd simultaneously embraces the two contradictory ideas that "AI don't work!" and "AI is taking all our job!".

Well, that’s a false dichotomy, isn’t it? Both statements can be true at the same time; indeed, even in the same company.

Article after article has shown companies ignoring the warning signs to chase the immediate returns AI promises, by letting go of workers, only to return to hiring for those same jobs AI promised it can replace because of the lag in realizing too late it didn’t live up to expectations over time.

Management and leadership are still starry eyed for those savings and automation that ultimately doesn’t play out as described.

Corporate America only relents when it hurts too much. It did this before AI in poor Agile and Waterfall projects or upcoming new tech that dazzled in unrealized gains and will continue to do so after AI’s dust settles. AI just makes larger promises than previous tech and we feel it far greater right now.

AI is a tool, not a miracle cure for replacing tons of staff. We’re still firmly in the “FA” part in the FAFO plan that is AI’s current market.

 

[BobHome]

"Some developers ... cannot understand it at all."
-- Just the way the AI wants it. /s

 

[Endymio]

Well, that’s a false dichotomy, isn’t it? Both statements can be true at the same time; indeed, even in the same company. Article after article has shown companies ignoring the warning signs to chase the immediate returns AI promises, by letting go of workers, only to return to hiring for those same job

Both statements cannot be true, no. If companies are firing people only to rehire them (a highly exaggerated statement, I believe, but let's assume otherwise) then the total number of jobs isn't changing, and AI isn't "stealing all our jobs". You can't have it both ways. Either AI is effective, or it isn't.

 

[GeoffreyA]

The more code I write and read the more I come to realize that comments the majority of the time are rather useless. Good code should be self documenting most of the time, by the time you're writing a comment you should ask yourself the question whether you shouldn't simply be splitting up functionality in more bitesize pieces and/or using more descriptive names.
Also greatly reduces the chances of things going 'out of sync' as it's much more intuitive to change the name of a method when it's functionality changes than it is to re-read comments above or sprinkled throughout it and update them.

Yes, good code should be largely self-documenting through simplicity, choice of names without being verbose, layout, etc.

In modern C++, there seems to be a bizarre effect happening, where the extensions to the language are leading to a shorter but more obscure syntax, at least for those familiar with the C++98-style idiom and earlier. Not to mention the colons everywhere cluttering the code.

 

[blakeem]

Developer of nearly 30 years. The issue is management mandating anything AI. One thing I learned over the years is that mandating or rewarding anything nearly always backfires. I've seen teams get reward for writing documentation and our docs became so full of junk and duplicates that they were useless and no one could actually use them anymore. I had another place that mandated 100% code coverage for tests, but the tests were terrible and just done to get around analysis.

With AI, my skills have actually increased because I'm learning more now. I'm against vibe coding and review everything and understand how something is because there is nearly always major issues with anything AI outputs. Vibe coding is the issue. I use AI to do testing so I have less issues from QA and UAT. I can use more strict linters and create higher quality unit tests than I'd otherwise have time for. AI is best used to augment and speed up your work, not replace it.