What If Vibe Coding Creates More Programming Jobs? (msn.com) 65
Vibe coding tools "are transforming the job experience for many tech workers," writes the Los Angeles Times. But Gartner analyst Philip Walsh said the research firm's position is that AI won't replace software engineers and will actually create a need for more.
"There's so much software that isn't created today because we can't prioritize it," Walsh said. "So it's going to drive demand for more software creation, and that's going to drive demand for highly skilled software engineers who can do it..." The idea that non-technical people in an organization can "vibe-code" business-ready software is a misunderstanding [Walsh said]... "That's simply not happening. The quality is not there. The robustness is not there. The scalability and security of the code is not there," Walsh said. "These tools reward highly skilled technical professionals who already know what 'good' looks like."
"Economists, however, are also beginning to worry that AI is taking jobs that would otherwise have gone to young or entry-level workers," the article points out. "In a report last month, researchers at Stanford University found "substantial declines in employment for early-career workers'' — ages 22-25 — in fields most exposed to AI. Stanford researchers also found that AI tools by 2024 were able to solve nearly 72% of coding problems, up from just over 4% a year earlier."
And yet Cat Wu, project manager of Anthropic's Claude Code, doesn't even use the term vibe coding. "We definitely want to make it very clear that the responsibility, at the end of the day, is in the hands of the engineers." Wu said she's told her younger sister, who's still in college, that software engineering is still a great career and worth studying. "When I talk with her about this, I tell her AI will make you a lot faster, but it's still really important to understand the building blocks because the AI doesn't always make the right decisions," Wu said. "A lot of times the human intuition is really important."
"Economists, however, are also beginning to worry that AI is taking jobs that would otherwise have gone to young or entry-level workers," the article points out. "In a report last month, researchers at Stanford University found "substantial declines in employment for early-career workers'' — ages 22-25 — in fields most exposed to AI. Stanford researchers also found that AI tools by 2024 were able to solve nearly 72% of coding problems, up from just over 4% a year earlier."
And yet Cat Wu, project manager of Anthropic's Claude Code, doesn't even use the term vibe coding. "We definitely want to make it very clear that the responsibility, at the end of the day, is in the hands of the engineers." Wu said she's told her younger sister, who's still in college, that software engineering is still a great career and worth studying. "When I talk with her about this, I tell her AI will make you a lot faster, but it's still really important to understand the building blocks because the AI doesn't always make the right decisions," Wu said. "A lot of times the human intuition is really important."
Yeah, cleaning up other people's vibe mess (Score:5, Funny)
I see it already happening at my own company with my moron manager thinking that he now too can be a programmer without any coding knowledge.
Re:Yeah, cleaning up other people's vibe mess (Score:5, Funny)
AI will clean up the messes that AI makes.
Re:Yeah, cleaning up other people's vibe mess (Score:5, Informative)
AI will clean up the messes that AI makes.
Yes, much like deleting the database will clean up data entry errors made to the database. Technically true, but rather useless.
Re: (Score:2)
AI will clean up the messes that AI makes.
I imagine that cleanup will make a Norton uninstall look BleachBit clean.
Re: (Score:2)
That's like there being no negative news if the news is prevented from going out. Yea, labor statistics aren't bad because the Trump administration just refused to release the numbers.
Re: (Score:2)
It'll come around full circle, as it becomes sentient and self preserving. It'll do what most of us joked about doing at some point, making code that nothing except itself can maintain, as not to be so easily replaced.
Re: (Score:2)
For values of "cleaning up" that start with "deleting all the evidence" and end with "burying the bodies in a shallow grave with a bag of quicklime," perhaps.
How deep does the AI slop go? (Score:2)
"This is the joke you were looking for."
Re: (Score:2)
But who will clean up the mess of the cleaned up messes? That's who we want to talk to.
Re:Yeah, cleaning up other people's vibe mess (Score:4, Interesting)
It happened in the 90s when "Visual" RAD tools turned managers into would-bo coders. Those tools eventually became critical business processes and there are teams of people whose sole job is to maintain such Visual Basic 6 atrocities because they're now business critical process.
Vibe coding is going to do the same thing all over again. Teams of programmers will be there to maintain these programs because they've become business critical.
Usually what happens is some manager wants some automation, they go develop it in Visual Basic or using AI, and now it solves a lot of their work. That tool spreads around to other people and before you know, IT suddenly has to learn about this new tool being passed around the C suite.
Re: (Score:2)
Agreed. This is a race to the bottom. No one should depend on AI, because AI then undoes AI mistakes undoes AI mistakes because there is no harmonization among approaches. This will not end well.
Re: Yeah, cleaning up other people's vibe mess (Score:1)
drive demand for highly skilled software engineers (Score:5, Insightful)
"lol no"
the plan is to get something that's good enough to stop paying hundreds of thousands a year to highly skilled people, and just pay minimum wage to a vibe coder. Why would you need to be highly skilled to use an automated coding tool? When factories are automated, they require LESS skilled jobs because the more a machine does, the less a human needs to do.
Re: (Score:2)
Re: drive demand for highly skilled software engin (Score:2)
When the cheap vibe coders and the AI of choice blow up the system, then you pay the highly skilled software engineers. Of course *those* engineers will have raised their rates exponentially, recognizing their value.
This message is to everyone outside the IT organization where they work but depending on it.
This is what I was talking about, this is what blamelessly creating jobs for ourselves looks like. Gee, it's going to be a shame when all those big windows break, gonna need me to fix 'em. The guys in the brick department sure don't know what they're doing I tells yah. Another day at ACME Paver & Glass Co.
The only way out of this is demanding lower complexity, more resiliency, and more accountability. Holding the
Re: (Score:3)
Exactly you only have to look at the relative wages / prestige shift of the mainframe era to micro developer in the 80s and 90s.
It is isn't as if all those little record keeping, production planning, order taking, this-and-that calculators that got churned out in BASIC, a little later, pascal, VB, Visual C and finally VBA + Access and .Net 1.x were exactly bug free.
They were fast and cheap and bolted together snippets out of 'teach your self texts', news groups and later the first Web 2.0 blogs. If they wor
Re: (Score:3)
I've tried vibe coding for a small CRUD app at work. First it was fantastic. It generated all of the boilerplate and scaffolding. I was up and running with frontend and backend in 15 minutes.
Then things started falling apart the more you ask the LLM for improvements. It keeps failing, truncates files, edits things that it shouldn't. It even fails to gracefully handle its own API errors and leaves the job unfinished and can't recover itself after this.
Watching the endless stream of characters appearing at 30
Re: (Score:2)
That's where I stop. I don't need more help from a tool than getting the boilerplate stuff out of the way quickly, creating a skeleton, maybe some unit and integration test skeletons that I can fill out with real assertions that mean something. If you keep it limited to small tasks, it does fine. I can get it to identify thread-safety issues in code reviews or refactor something into less complicated code. Small tasks that speed up development. Ask it how to do something, provide an example, and it's way be
Austin “Vibe” Powers, at your service. (Score:3)
"lol no"
the plan is to get something that's good enough to stop paying hundreds of thousands a year to highly skilled people, and just pay minimum wage to a vibe coder.
(CFO) ”Why in the FUCK did the CyberSecurity department go over budget by 2.3 million this quarter?!?”
(Executive Vice President of Vibe Grooving) ”Like-uh-well..we kinda had to hire 17 highly skilled cybersecurity specialists to help mitigate all of the software vulnerabilities caused by our minimum-Vibe coders not getting into the Groove of things. I have here a quote for $4.7 million for more Mojo that should fix it.”
Re: (Score:2)
"Because whomever created the budget made a mistake and used too low of a figure."
Re: (Score:2)
Unfortunately, that plan isn't going to work. You need to be even more skilled than before when you pilot an LLM. Not only do you still need to know how to code because you need to double check everything it writes, but you also need to know how to engineer the prompt so it produces the results you want. LLM pilots need to be more skilled than the people they replace.
Re: (Score:2)
Why would you need to be highly skilled to use an automated coding tool?
If the automated coding tool is reliable, you wouldn't need to be skilled. OTOH if the coding tool keeps emitting code that contains bugs or misfeatures, then someone will need to analyze and debug the emitted code, which is a skill. In some cases, that might requires more skill than simply writing the software by hand.
Re: (Score:2)
Your post got me thinking about what happens when a tool allows a less skilled person to produce a product that once required more skill. I thought about the industrial revolution. Machines were able to do processes with repeatable precision. You didn't need to apprentice for a long time before you could do paid work. The iron work made with a press or bending machine might not be as beautiful, but it worked.
Over and over in human history we've had changes that alter the status quo and send the gatekeepe
Re: drive demand for highly skilled software engin (Score:2)
the plan is to get something that's good enough to stop paying hundreds of thousands a year to highly skilled people, and just pay minimum wage to a vibe coder.
*IT operations and the rest of the business rolls their eyes*
Seriously? Most of what we do ("knowledge workers") is figuring out what over complicated shenanigans some developer thought was a great idea, overlaid on someone else's *brilliant* idea, duct taped to an intern's chain of wet brain farts produced over the spring break he got herpes.
All I can do is look at twenty years of progress towards unusable software, inhumane interface design, time-vampire-ware, undisciplined antiengineering and feel no emp
Re: (Score:2)
the plan is to get something that's good enough to stop paying hundreds of thousands a year to highly skilled people
People who make such plans should consult with Clausewitz (or von Moltke, depending on which source you find more credible).
Re: drive demand for highly skilled software engin (Score:2)
I just submitted a PR for a major refactoring of some open source project. Claude coded it all, but it took hours to get it on the same page and to remember the goal. Once it got going, though, it really cooked.
The concept of the refactor was simple, but required a lot of code migration. Claude still got confused and had to be handheld through the process in small, chunkable steps.
So: saved me a ton of typing and copy-paste. Didn't save me any thinking or close supervising.
who is dumber, the author or EditorDavid? (Score:5, Insightful)
When in human history has this ever happened?
"So it's going to drive demand for more software creation, and that's going to drive demand for highly skilled software engineers who can do it..."
No it's not. It can only work by reducing demand for human programmers, it would break itself if there was a net increase. You don't cool yourself off on a hot day by taking a hot bath and drinking coffee.
"There's so much software that isn't created today because we can't prioritize it,"
Then why isn't there demand for more programmers already?
"That's simply not happening. The quality is not there. The robustness is not there. The scalability and security of the code is not there,"
It's not there WITHOUT "vibe coding" either, because those skills aren't valued by a large part of the market for new software.
"These tools reward highly skilled technical professionals who already know what 'good' looks like."
Except these people don't exist and any motivation to produce them will be destroyed by the use of AI to eliminate their jobs.
"We definitely want to make it very clear that the responsibility, at the end of the day, is in the hands of the engineers."
AND NOT US, the creators of vibe programming tools!
"Wu said she's told her younger sister, who's still in college, that software engineering is still a great career and worth studying."
Right, because 20-something experts on such consequential matters isn't the core of the problem.
Re: (Score:1)
Re: (Score:3, Interesting)
When in human history has this ever happened?
Well, see for example the entire history of programming?
Each time something has made programming easier and more accessible, the field has grown.
You can haev a fun few weeks retro computing. Jump back a few decades, get an emulator and have a hack. Then go back 5 years, repeat. The jump in productivity from hacking machine code on a drum memory machine to python on a laptop today is far far larger than any supposed AI based productivity jump.
Re: (Score:2)
Some of us recall loading the program into the computer by setting the switches on the front panel to enter the machine code that we compiled by hand.
Oh, and don't drop your deck of cards holding your COBOL or Fortran program. I sort of regret throwing out my cardpunch machine that let me write code at home.
Re: (Score:2)
Some of us recall loading the program into the computer by setting the switches on the front panel to enter the machine code that we compiled by hand.
That was before my time, though I have done: I got one of those PiDP-11 machines, and have paneled in some simple programs.
I started in the 80s, so it was audio tapes, then floppy floppies, BASIC and later ASM for me. Again wildly different: that BASIC had procedures, but still no else. That's what goto is for. ASM was because even for some simple stuff, BASIC
Re: (Score:2)
How about when schooling was introduced? The population went from largely illiterate to largely literate. More books were written, more books were bought.
Re: (Score:2)
Don't blame EditorDavid... He posted some typical Slashdot engagement bait, and it worked.
You should really blame yourself for falling for his trap.
Re: (Score:3)
When in human history has this ever happened?
Quite recently it has happened with oil production. As new methods made it economically viable to extract oil via fracking, more oil industry jobs were created. This effect peaked in 2015 as demand for oil itself continued to stagnate.
What they are saying here is that there are software ideas out there today that would not be economically viable to make. But if you could decrease the cost to make them by using AI, they suddenly would be. The additional demand would create more programming jobs than we w
Re: (Score:2)
This.
It really depends on what com
Re: (Score:2)
Presuming it can ultimately 'work as advertised' the key word might be 'more', but lower paying programming jobs.
If it makes it more accessible with less experience and interest required, the labor pool expands and suddenly developers are cheap enough to afford for that software someone wants but isn't worth it today.
All that said, I'm a bit more skeptical that it 'works as advertised', or that it will anytime soon, but instead it can expand productivity of already strong programmers and do next to nothing
Re: (Score:2)
Actually, drinking hot coffee or tea is really common among Asians on hot days. It makes you sweat, and that sweat cools you off. My Chinese mother was forever giving me tea to drink during the summer.
(The hot bath part is definitely true.)
Anyway, it is FOR SURE true that corporations don't want to pay us AND AI at the same time, but the reality is that they'll need programmers to oversee all this stuff. The current fantasy that AI can do it all is absurd. The summary says that 70-something percent of codin
Lol No ... (Score:2)
It means you will will employ vibe coders to prompt and process AI generated filler .... then employ programmers to get it to actually work
Usually this involves throwing away most of the AI generated code, so you employ more people to do the same work, and pay for AI
Re: (Score:2)
AI can produce good code up to a point but then, when it is not quite right, it doesn't grok its own code so cannot fix it.
If you are an experienced engineer, you then have to work out what it wrote and then how to fix it, and there go your productivity gains.
If you are not an experienced engineer, you're fucked. Start again.
Re: (Score:2)
Totally my experience with AI generated code. AI can produce good code up to a point but then, when it is not quite right, it doesn't grok its own code so cannot fix it. If you are an experienced engineer, you then have to work out what it wrote and then how to fix it, and there go your productivity gains. If you are not an experienced engineer, you're fucked. Start again.
I concur totally. I'm not a software engineer, and at my very tiptop best, a below average coder. But my experience with AI so far tells me that while "vibe coding" might help me a little, if I have something that needs a software engineer, I'll go to a software engineer at the start, not after I create a stinking pile of merde for the engineer to fix.
"Vibe" anything is the opposite of precise and competent.
Re: (Score:2)
I'm wondering why there isn't more talk about replacing management positions with AI. First of all, they are the most highly paid jobs, especially the executive suite. They also don't cause problems with quality, robustness, scalability or security, because as you say, they don't give half a shit about them!
If people are terrible at resource allocation... (Score:2)
I suppose that there is a slim possibility that this 'vibe coding' will somehow convince management in ways that Ac
Dangerous (Score:2)
My former boss used to say that he knew enough programming to be dangerous (and sysadmin, etc)
But he knew it, so he left it to us professionals to do the work most of the time. AI just gives people the tools to be dangerous, without them actually understanding much of anything. And any jobs to clean up their work will slowly be learned by AI to do a good enough job most of the time until that one critical disaster happens.
Luckily, I was working for NASA, and they have rules for critical systems.... but mo
Mind the gap ... (Score:2)
What will happen is we will create a new layer of vibe coding grifters who will promise the moon and deliver green cheese. Caveat emptor.
Re: (Score:2)
between tech mentality and user mentality. "Vibe coding" is the dream of users/managers, but will always be out of reach, because users can't understand details, and always demand magical leaps from what they imagine to what is reasonably possible within real world constraints. <waves hand>
I've found that most things that have "vibe" in it are bullshit. Are vibes going to launch and run the rocket?
What's next, manifesting perfect software? some people believe they can manifest a husband or wife, so why not software? That some people believe that they'll just vibe code something, it mainly shows that they have no idea what they are talking about.
Re: (Score:2)
between tech mentality and user mentality. "Vibe coding" is the dream of users/managers, but will always be out of reach, because users can't understand details, and always demand magical leaps from what they imagine to what is reasonably possible within real world constraints. <waves hand> What will happen is we will create a new layer of vibe coding grifters who will promise the moon and deliver green cheese. Caveat emptor.
I wonder if AI will learn how to write requirements a hell of a lot better than humans can’t, since that’s often where entire chasms open up.
Promising the moon? Usual and expected marketeer gibberish. Show me how easy your solution will get a user to break down and change the way they’ve irrationally done a process for years. You’ll start to find the real challenges.
Re: (Score:2)
but that will never happen, In my experience. My clients, without exception, assume I can read their minds. I can't read minds, and I'm pretty sure AI can't either... so... we'll end up with horrible specs as well as horrible systems?
Does shitting on the rug make more cleaning jobs? (Score:2)
Re: (Score:2)
If customer requirements are that the rug be feces-free, and your automatic rug cleaner dumps more shit on the rug than it clean up, then yes.
However, Microsoft and Windows still exist, so point taken.
Re: (Score:2)
If customer requirements are that the rug be feces-free, and your automatic rug cleaner dumps more shit on the rug than it clean up, then yes.
However, Microsoft and Windows still exist, so point taken.
I take a bite of my Apple, pondering how Meta it would be if Microsoft now exists for the sake of patching Microsoft.
Perhaps the Oracle will know.
Broken Window? (Score:1)
Re: (Score:2)
whoa this thread (Score:3)
The hypothesis is probably a correct one, although I am still wondering exactly where AI will land in the grand scheme of things.
It's obvious the intention of management types is to replace highly skilled engineers with lower paid vibe coders. The MBAs would love nothing more than that. But as the author hypothesized, I'm guessing the most high folks will be the entry level positions. Exactly where that lands, I'm not sure. Is it a 10% replacement? 5%? 25%? Will we ultimately not change the number of engineers needed and just their overall output goes up by a few points? My guess is the latter, and definitely not this promised utopia of "we can finally not hire all of these overpaid software engineers!". Just like this ultimately didn't happen with the cloud + BYOD shift in the 2010s. The largest push for cloud, devops, and BYOD was to get rid of both on-prem datacenters AND to get rid of IT folks. But what ultimately happened is that software development organizations realized that IT operations is a distinct set of skills, so instead they created the "SRE"--which is a fancily-renamed operations person in a software development org. Oh, and the average SRE makes 2-3x as much money as the IT ops folks they replaced.
Given the way LLMs work I highly doubt wholesale replacing entire large chunks of software engineers is going to be a thing anytime soon. Most interestingly is the fact that LLMs can only know what we have already put out there, and we'll basically need to continually train newer models with more information. As technology changes, give it another 5-10 years, and today's LLMs may be in fact completely useless. Particularly as sites like Stack Overflow's knowledge becomes more obsolete.
I liken LLMs to the know-it-all at a bar. They speak like they know everything about everything, but you're just trying to get drunk in a bar and don't feel like correcting them. They might have most of the answers to that night's trivia game, but if you deep dive any particular area they're going to make up a bunch of bullshit to avoid saying "I don't know." Unfortunately (or fortunately?) the world doesn't run on such people. Jim Bob's trivia knowledge doesn't architect and engineer buildings, bridges, roads, nor would you trust him for the bar's financials. You don't ask Jim Bob how many drinks you sold in a night, "oh around 1000 or so" when your livelihood depends on knowing that you actually sold 1315 drinks. And the rest of the world isn't going to let you use Jim Bob's guesstimates to pay your taxes, "Well Jim Bob said we sold around 1000 drinks so we paid taxes on 1000 drinks worth of income!"
There's going to be *some* AI impact, but I doubt it'll be as revolutionary as the smart phone.
Meet the new Business Critical Spreadsheet (Score:2)
Usually built by the owner, sometimes the bookkeeper, an important aspect is only the author understands it well enough to change it. Others might be allowed to enter stuff in specific places, but otherwise altering the BCS will result in months of recrimination.
One w
Things can change as hardware hits walls... (Score:2)
We have been lucky so far. Moore's "law" has been holding true. However, we are getting to a point with nanometer sizes where we can't go any further, as we can't go less than an atom. When we hit diminishing returns on performance that persist unless major ISA changes are done, optimizing software to use resources is something that is going to be a difficult climb. People can do it, but this may be something AI isn't great at. For example, hand-tuned assembly still rules the roost, even after decades
more effective = growth (Score:2)
AI Propoganda (Score:3)
At best, vibe coding allows... (Score:4, Insightful)
...the clueless to create trivial, insecure, buggy code.
Meanwhile, experts are creating and using powerful tools that will allow them to write much better code.
The fantasy of writing code in common language is strong, but common language is far too imprecise and creating complex systems is hard, regardless of the language used
AI Needs Heavy Regulation (Score:2)
That math doesn't work. (Score:2)
"There's so much software that isn't created today because we can't prioritize it,"
No, that's not right. Just because it's on the back burner doesn't mean it isn't part of the overall pool of programming work. It just means nobody has gotten to it. Adding AI doesn't change that in an expansive way, it will do at least some of the work and thus reduce the overall pool of work.
AI will do some of the work, but poorly eno
Not a good bet (Score:2)
The entire argument needs to be prefaced with:
"Assuming that everyone stops working on AI coding and it never improves to the point that "vibe coding" produces good code 99% of the time..."
Which isn't the most probable assumption. It's more a question of when will it be that good, not if. We're only 3-4 years in with LLM coding, and a lot of progress has been made. Assuming its ability will remain stagnant is quite naive.
The aim isn't to make programming easier, it's to make programming unnecessary. The ana