Vibe coding will deliver a wonderful proliferation of personalized software

For most of the last year, the phrase 'vibe coding' seemed more punchline than possibility. That outlook altered significantly over the last month after step-changes in quality mean vibe coding tools now generate code that’s good enough to rewrite expectations about how IT will operate before the end of this decade. Until …

[Michael Hoffmann]

This article on El Reg is the equivalent of some ancient Christian walking into the arena, knowing the lions are already waiting and blowing a raspberry to the Emperor's Seat.

But I'll bite:

What you're saying is that an experienced, senior developer can indeed make use of an LLM to steer it to some code that might actually do what you want it to. Simply assuming that you are one. Though making the bold claim that your app is "professional level" had me go "orly?"

I'll leave aside for now how you checked the code wasn't a dismal mess, whether you had tests (unit and integration) generated and verified they weren't just a giant set of methods mocking everything and returning "assert True"

How will this aid an inexperienced coder (or worse, a Dunning Kruger manager)? Is the success criteria "well it seems to do what I want and if I squint the results look correct"?

I've done what you have and frankly, I don't call it "vibe coding" because "vibes" don't come into it. I have an idea, some very precise prompt instructions that can on go on like a Tolstoy novel, and then expect to be doing a lot of hand-holding. To the point where I wonder I shouldn't have just written the whole mess from scratch without boiling the planet.

[ParlezVousFranglais]

100% this - just because a bunch of clowns can hammer together a red bull soapbox to race down a hill doesn't mean they can design a real car, and that is proved by the destruction most of them leave in their wake.

Vibe coding is the IT equivalent - if you want to have a bit of a muck around, then go for it and have a laugh - but don't expect to design anything serious or dependable in any way because it will end in bumps, bruises and very likely complete disaster at some point down the road...

[LucreLout]

It's not even that good, I think.

Your soapbox racers at least know what a car is. They also know how to rebuild what they built.

AI is non-deterministic, so future models or even an older version of the same model, will all produce different results. That is going to make maintaining the vibe coded app far more difficult than folks are pretending. And we know for sure that 80 of the work with software is maintaining it after V1.

It's useful to pros in certain circumstances, the problem is the buffoons that think makes them the pro. A junior dev with AI is still a junior dev, it's just they're cranking out a lot more junior stuff for a senior to rework or fix. Civilians have no chance - they won't know what's missing.

[Eric 9001]

LLMs are in fact deterministic - it's just that randomness is added via a temperature variable, so the chatbot doesn't output the same thing every time the same input text is inserted.

Although it's true that future or past models have completely different outputs.

Haphazardly copying parts of badly written software from online with a LLM, may produce the result of a program that compiles and seems to work, quickly, but if you actually want a program that works correctly, you then need to go and fix the many wrong things in the mess (which is slower than just writing or copying software without a LLM - but the LLM user thinks they did it faster).

[TaabuTheCat]

So, so true

The non-deterministic part. I saw it firsthand a couple of weeks ago when Gemini changed models. I asked it to add a simple new piece of functionality to code it had already written with the prior mode, add it to the simple procedural program it had already written, and it completely re-wrote the program to be a class-based with one class and all the procedural stuff defined as methods. It then created an instance of the new class and called all the methods, one after another.

And don't get me started on its ability to break working code. Gemini makes a mistake, I correct it, it fixes the mistake and then some number of iterations later, as I'm asking for new functionality, it puts the old mistake right back in. Maddening. And as the OP noted, by the time I was done I was sorry I just didn't write it myself. At least It would maintainable. Now whenever I ask for a change I have no idea what's going to happen.

[Not Yb]

Re: So, so true

It put the old mistake back in, probably because the old mistake was still present in the prompt history. Going back and editing the prompt to incorporate the fix might be a good idea, but that does ruin the 'vibe' part, doesn't it?

[TaabuTheCat]

Re: So, so true

I had Gemini correct the mistake with a prompt, so the history was there of both the mistake and the fix.

[jake]

"I don't call it "vibe coding" because "vibes" don't come into it."

I've taken to calling it Jive Coding, for somewhat obvious reasons, some of which you touched on.

_{Strangely, the folks who have taken to this new scam instantly understand what I mean, and hate me for it. C'est la vie.}

[ecofeco]

Jive talking

You're telling me lies

Jive talking

You wear a disguise

Jive coding indeed. Excellent moniker!

[David 132]

"Oh stewardess, I speak Jive..."

[PRR]

In case anybody missed it: https://www.tiktok.com/@richschneider/video/7255704963207138602

[EricB123]

French and vibe coding

I got down voted for using the French expression "C'est la vie". Oh wait. I wasn't replying to a vibe coding article.

[katrinab]

Re: French and vibe coding

If we are talking French, my favourite French phrase is «Chat j'ai pété»

[Anonymous Coward]

Do you smell Tulips !!!???

This ^^^^^^^ 10^10^10^10^10^10^10 times at least.

If I was a great and talented artist I could produce great art with an A3 piece of paper and a selection of coloured Crayons.

I would NOT expect Average Joe/Joanna to be able to do this !!!

The pitch for 'AI' is always 'Would' do this ... never 'Could' if you have the right skills to start with.

'AI' is too random and lacking in control to be unleashed on the world ... You need tools to work 100% of the time.

Warnings and listing of exceptions does not work in the 'Real world' ... people will use the 'AI' expect 'right answers' and impose the randomness on the rest of the world.

I do not want 'AI' generated ANYTHING if it is not checked for completeness/validity/overall safety by SKILLED people not another 'AI'.

The continuing push to find some arena where 'AI' is accepted & used to generate profits and start the general acceptance into our everyday lives is beyond desperation.

It is not fit for use ... period !!!

Unless it improves immensely not just in quality but repeatability across all possible arenas, then 'AI' is just a clever trick !!!

If 'AI' is involved in the production of things that impact people & lives in general then this scam cannot be allowed to continue.

Real harm WILL happen as a consequence of ignoring the limitations of the current 'AI' ... real harm that will cost real money (as money is the only thing that talks when referencing 'AI') !!!

The bubble cannot burst soon enough !!!

:)

[drankinatty]

Re: Do you smell Tulips !!!???

"The continuing push to find some arena where 'AI' is accepted & used to generate profits and start the general acceptance into our everyday lives is beyond desperation."

That's the part that really worries me. The AI companies are financed with $2T of circular debt secured by the promise of non-existing future profit. To the point they floated the idea of duping the US taxpayer into backing the debt. The echos of "credit default swaps" are loud indeed.

While the models can spit out code that compiles and functions, absent continual revision by human experts, putting any of it into production is a fools errand. When very experienced human can argue of the meaning of text in a language standard, those areas of interpretation are an LLM's Achilles's heel to ever being able to get things 100% right. Whether behavior is defined or not hangs in the balance. That bodes well for continued employment of experienced coders.

[HereIAmJH]

Job losses

What you're saying is that an experienced, senior developer can indeed make use of an LLM to steer it to some code that might actually do what you want it to.

I think we're probably going to have to learn to walk before we can run. The example in the article was too complicated. I don't think we should be trying to have it flesh out complete apps yet. Instead, that senior developer is going to feed the LLM the various stories in the application. Maybe down to button click level. What if you could feed it a unit test and have it build an implementation? It won't have inexperienced coders (or business analysts) building apps right now. But it will allow corporations the ability to gut project teams. You won't get rid of the highest paid developer yet, but you'll dump 5-10 junior developers for each one of them. That's a significant savings already.

And of course, as we use the tools the AI designers will learn the deficiencies and the senior developers will work themselves out of a job. Kind of like training your off-shore replacement.

[tfewster]

Re: Job losses

And how do you become "an experienced, senior developer" if there are no junior developer roles?

[HereIAmJH]

Re: Job losses

How do you learn to make buggy whips? If AI coding becomes what they want it to be, software developers might as well be blacksmiths. Developers can sit and talk about the good old days with the airline pilots and taxi drivers.

[O'Reg Inalsin]

Re: Job losses

"If ...". These premature victory declarations reminds me of Bush declaring "Mission Accomplished" 6 weeks into the Iraq War.

All we have to do now is build data centers as far as the eye can see and the rest will follow, right?

[HereIAmJH]

Re: Job losses

With "Mission Accomplished" you are confusing technical limitation with a political reality. Technically, it would have been possible to do it in a week with WWII technology if it was politically acceptable to have high civilian casualties.

From a technical perspective, when was the last time you had to tune a carburetor on a car? OBDII has turned many automotive mechanics into parts changers. While I don't want to be anywhere near a Waymo, it's not logical to think that self driving cars aren't the future. The things we do with the Internet and cell phones was unthinkable 25 years ago. Some technology fails, but never bet against technology in the long run.

And it's not build datacenters and the rest will follow. AI is driving the construction of datacenters, new energy production, advances in computing via GPU/NPU. Right now businesses are willing to invest in technology that will change the landscape of IT. The degree of change is dependent on how much money they are willing to spend.

[Anonymous Coward]

Re: Job losses

"The degree of change is dependent on how much money they are willing to spend."

This is an example of the mindset that the 'AI' techbros want EVERYONE to accept as true and valid.

It is not close to reality at all !!!

It is the usual 'hope against reality' mindset.

We are experiencing the reality that spending huge ammounts of money does NOT make technology work .. there is not a point where the poundage of 'money spent' tips the scale of technology and the problems you had at $1Billion are vanquished by spending $10Billion.

The problems of 'AI' that means it is not working for everyone 100% of the time are not solved by money BUT solved by understsnding why current 'AI' fails and what you need to do to make it work ALWAYS for ALL problems.

Many of the evangelist for 'AI' reach the point in their argument where they state that all the problems will be solved in the near future and we should accept it is coming so stop fighting the inevitable.

Nowhere is the solution to our failing 'AI' enumerated and it is simply a matter of 'Faith' because that is what 'They' want !!!

This is the 'Hope against reality' point and it is so at odds with the huge focus on real actual technology that has got us here.

I have been hoping for Flying cars for 50 years ... my hope has not solved the problem and they are still NOT here !!!

All the people who are so grounded in the deepest technology are coming down to 'Faith' & 'Hope' for the answers and the reason we are supposed to ignore what we see in favour of what others wish for !!!

I am in an industry where we know that writing code that states 'Here is where the working code goes' does not compile to anything useful !!!

The same people surely cannot be expected to believe that having an effective placeholder that states 'Here is where the working 'AI' will be' is any more believable ... even if you hope it is soon !!!

'AI' is a scam that is hoping that it will be delivered BEFORE the money runs out and/or the customers realise the wait is never going to end.

There is no evidence that the 'Working 'AI'' is just around the corner ... no one is saying it is almost there ... what we are being sold is something that gives an indication of where true 'AI' could go IF we just keep waiting and accept an almost working version for now (for a very small value of almost).

'AI' is being sold in a similar manner to how every grifter has ever worked since the 1920's.

A good story and a variation on 'get in now before the price goes up and your future profits go down' !!!

FOMO is making a small number of people very rich on the back of idiots !!!

Quick ... stop listening to your brain and listen to the convincing story and imagine the mountains of cash you will make !!!

:)

[Mark 65]

Re: Job losses

Well, 'Faith' & 'Hope' best get vibe code because we’re all waiting on those sisters to get the job done

[doublelayer]

Re: Job losses

I think the Iraq War example is a lot closer to the mark than you claim and has similar restrictions and answers as the AI example, starting with a simple problem with your statement: "Technically, it would have been possible to do it in a week with WWII technology if it was politically acceptable to have high civilian casualties."

Fine, everything's on the table. You can carpet bomb, small nukes, kill anyone you want. What can you actually do within a week? If the goal was simple, find where Saddam Hussein is and kill him, maybe you could. Even then, all sorts of things could get in the way of that quick a timeline. That's a short timeline for finding a hiding place if he was fortunate enough to go to one. That's analogous to code that you only need to work once and can go wrong in many ways, and we do have plenty like that. If I can write something quick and dirty which will get the results I want rapidly, even if I cause several errors using the run first and check whether I followed the API requirements later, I can do that too.

If your goal is a little bigger, obtaining the surrender of Iraq's armed forces, now it's trickier. Oh, it might look easier because somebody's going to surrender when you're threatening to nuke Baghdad, but you don't just need them, you need everything. That's analogous to something quick but that you're going to run multiple times. Little bugs are no longer acceptable, but at least there shouldn't be so many you need to iron out because the task itself is small.

But neither of those was the complete goal, hence why the actual war was long. Debating the goal of the attackers is not relevant here, but no matter what you think it was, whether that's making Iraq a democratic country, eliminating terrorist groups resident there, preventing more genocide, weakening other countries' diplomatic or military positions without going to war with them as well, those are not simple goals and no level of weaponry was going to make that happen because the trouble was not easily locatable and bombable. Getting that after World War II took a lot of time, money, and some threatening countries that the allies were protecting the vanquished from, and repeating that is not easy either to achieve or to convince politicians and citizens to pay for. And that's what people who write software mostly need as well. They need software which continues to work long-term, that can be fixed or improved without rewriting big chunks, reasonable confidence that those little changes won't introduce massive new bugs, testing to confirm that, and correct documentation and understanding from those maintaining it. These are not simple requirements and current LLMs do not do them at all. Nor is it likely that just throwing more money at it will make that happen any more than more bombs in 2003 would have prevented ISIS from forming. There are some problems that can be solved by throwing more money at them, but diplomacy and inventing unproven technology tend not to be among them.

[Not Yb]

Re: Job losses

Exactly, thinking that any real-world war ends with anything other than humans dead and ravens fed is fundamentally misunderstanding the nature of war. "We have goals" doesn't matter so much when reaching those goals requires people to die en masse.

[Anonymous Coward]

Re: Job losses

Considering that the reason for starting that war was bullshit, he could have made that declaration early on the second day and still been 100% accurate and correct; however, that wouldn't have guaranteed a large purchase order to restock military supplies.

[martinusher]

Re: Job losses

The 'buggy whip' simile is overused.

The key notion in this article is "Domain specific knowledge". I've always told people that knowing how to write code is like learning to write English. Its an essential skill that's needed to communicate effectively but being able to write well doesn't make you a great author. (Although you might be able to simulate being an author.) I've always programmed but I don't think of myself as a programmer, I'm primarily an engineer, I just happen to use software to implement my designs.

[HereIAmJH]

Re: Job losses

Knowing the business is definitely an asset, but I've worked with a LOT of short term contractors that know the software tech and absolutely nothing about the business. It happens a lot in large companies that want to tackle a large project but don't want to increase headcount. I have worked on multi-million dollar projects where 90+% of the team were contractors.

If the AI is doing the coding, then can a business analyst do the guiding? At smaller companies can power users interface with LLMs to build the apps? Let's see how much shadow IT comes out of this.

From the downvotes I get every time I bring up AI killing developer jobs, a lot of people here don't like the idea. I have decades of experience writing code, I like doing it, but I wouldn't start a career doing it now.

[Anonymous Coward]

Re: Job losses

AC obv.

I'm aware of analysts who create a spec, then use one AI to refine it & produce a set of prompts suitable for the development AI (a different AI system) tool to use.

,, and then proper devs have to kick what is produced into shape (or touch, depending how dire it is)

[LucreLout]

Re: Job losses

Sadly, I don't think "you" will.

By the time today's junior gets through ten years to be vaguely senior, the ai will already be doing senior level work.

This all, of course, assumes we don't hit a ceiling. I mean, every actual senior developer knows the easy 80% of the project takes 20% of the time, the hardest 20% the rest of it. Maybe the pace of advancement in coding will slow.

[EricB123]

Re: Job losses

Well, you start by taking out an enormous student loan, and then hope you beat the enormous odds of finding a junior level position.

[nijam]

Re: Job losses

> ...we're probably going to have to learn to walk before we can run...

Whereas out there, AI is running before it can even crawl.

[that one in the corner]

Re: Job losses

That is how the ill-equipped always tackle a steep descent: just start walking down without seriously examining the problem (and finding out what has happened to other people trying the same thing), instead of crawling and crabbing carefully across the landscape.

Soon you are running faster and faster, managing to stay upright and congratulating yourself on how well you are doing, you are a natural at this, fie to all the naysayers.

Now you are out of control, the momentum is too great to let you go in any other direction than straight down in the grip of a force that keeps you accelerating. Your feet are hitting the ground so hard the scree is flying out in all directions; over there a flying rock slashes an innocent bystander and in the other direction it has created a landslide, undermining the plateau from where you started, those same inevitable forces causing the previously solid ground to collapse beneath the onlookers above.

The bottom is a sudden jarring moment of realisation that you've just run out of luck. The best you can hope for is a quick end; otherwise you will lie there immobile, nothing to do but watch the lives of those you've affected come crashing alongside.

PS

In the above, for "you" please read "LLM tech bros"; putting that in every time would have ruined the flow.

[teebie]

Re: Job losses

Crawling may be the only thing it does well

[Anonymous Coward]

Re: Job losses

Sorry, "stories"? I think you get to spend half of your time in meetings. The other half of your side are probably thinking of new and inventive words to replace well-known ones like 'loop' because moving to a quasi-rope motif will insure that a large number of halfwits will find offence in the word 'loop'.

[HereIAmJH]

Re: Job losses

My last employer went agile. Of course we spent a stupid amount of time in meetings. I think on one project we had 6 PMs of different flavors with only 20-30 people actually writing code. Hour long 'standups'. God help you if you interfaced with multiple teams. As a senior developer I spent 3-4 hours a day in meetings of some kind. Forget about doing any coding when audit season came around.

[Brewster's Angle Grinder]

"Though making the bold claim that your app is "professional level" had me go "orly?""

From the article, "I wrote a few VRML 1.0 browsers thirty years ago, so I came to this with deep domain knowledge."

You may not like his opinion. But I suspect Mark has a clue about what makes it professional level. At least engage in the argument in good faith.

[m4r35n357]

Prolly best not to tell people what to do!

[doublelayer]

I don't have a position on it either way, but those two statements come from the same source. Someone who doubts the first might also doubt the second, with the option that the author is exaggerating or inventing his history, that he has forgotten important parts of that history, or that, while having done exactly what he claimed in the past, he still doesn't have enough understanding or testing of his new tool to be so confident in its professionalism. Among other things, his claimed history with VRML could mean that the software is professional in its ability to do the tasks he's testing but doesn't necessarily prove that the code is professional, an important aspect if it is to be later maintained. Any of these things would have to be supposition without external knowledge of the author's skills and experience which I certainly do not have, but so would any assumptions that those statements are both true.

[steviesteveo]

Certainly the implication is that there's nothing else to point to in the last 30 years. I'd really expect Reg readers to have the professional nous to read between the lines about the wider CV here

[Roland6]

>” But I suspect Mark has a clue about what makes it professional level.”

Having domain knowledge is helpful and this does see a reasonable test of both the capabilities of and what is actually involved with using “Vibe” (*) tools.

However, Mark’s comment:

” VRML 1.0 translation tools have been lost in the mists of time. There's no way to load a VRML 1.0 world and view it.”

Is easily discoverable through a simple Google search to be found to be incorrect, even for those using a Mac there are commercial tools (£) , open source projects and free tools…

I’m also a little surprised Mark, given his CV claims doesn’t have the source code to any of the tools he developed. In the context of the evaluation, this code, which almost certainly would not be on any public repository, hence enabling it to be used for code comparison.

(*) I use “ Vibe” to mean any AI software development tool.

[tiggity]

@Roland6

"I’m also a little surprised Mark, given his CV claims doesn’t have the source code to any of the tools he developed. In the context of the evaluation, this code, which almost certainly would not be on any public repository, hence enabling it to be used for code comparison."

I do not have any of the source code for stuff I have written for employers - (fairly standard UK IT employment contracts will always give employer rights to the code you produce for them - some try & claim rights to stuff you produce at home doing hobby stuff too, but they can FO on that aspect). A competent employer (half decent security measure / auditing) will ensure you cannot take source code away with you. *

There is some of code in open source projects (ironically, not committed by me, all based on stuff I helpfully submitted on mailing lists to solve problems people had** & people on those lists thought it good enough to add to the repo, giving me a credit when pushing the commit.

* Yes, there are always further loopholes, even when the obvious email, FTP loopholes etc. are closed (port closed central) & all bar a few websites / IP addresses blocked e.g. unless you work somewhere properly secure, then mobile phones are ubiquitous in the workplace, so nothing to stop you taking surreptitious photos of code on screen, but how many people CBA to do that for code they wrote at work?

** Even more irony, most of those code snippets are in languages that are not in my true core competencies, I just often have a good knack for problem solving (& for "non core" languages the far harder part is implementing that solution in the required language!)

[Roland6]

@tiggity

Don’t disagree.

I was working on the basis that having written “the book” and provided a toolkit:

” Written by the creator of the VRML graphical standard, this book shows users how to create, display, view and manipulate VRML objects on the Internet. An accompanying CD-ROM includes a collection of VRML objects, VRML converter and object-viewing software.”

That he would have independently authored various tools including a viewer.

Looking at Mark’s publishing history, I suggest he has a new book in the works that is focusing on the use of Claude et al. If this article is a taster, I expect it will be an interesting and informative read.

[takno]

I focused a bit on the "I was given $300 in coding credits", along with the fact that the author still effectively spent a couple of days of work on it. That doesn't seem to make it particularly cheap to me - $1-2k (depending on day rate) for something that is almost certainly already available as Open Source or COTS software.

The author also appears to have donkeys years of code and domain experience, so really isn't Joe Public, and isn't really qualified to vibe code. The problem itself is ancient, with likely a huge amount of documentation and sample code available, so it's really the perfect problem for an LLM to work on, and not especially representative of most problems.

As it happens I did use Claude Code to put together some python scripts over the weekend myself. Again I'm an experienced coder and it's a well-trodden domain in Python. It probably saved me about 3 hours of code-writing, at the cost of around 2 hours spent trying to work with its terrible advice on packaging. The scripts did an _okay_ job, and since I don't really like writing in Python, I'd regard the fact that I didn't have to do any as a small win.

It's an okay job for some tasks, at least while the free credits last. I don't see it being as impactful as even the advent of reasonable IDEs though.

[Rahbut]

Things like Claude, OpenCode and Antigravity are quite impressive. They can do a fair bit really well when used correctly by the right people. And they're getting better.

None of that is contentious.

But let's be honest, they're best at things that have been done a gazillion times before - the average CRUD app in the enterprise is surprisingly easy to churn out, because that kind of work is fairly formulaic. Whether we like it or not, not every developer is working on something as complex as the control plane for the Large Hadron Collider, even if some will make it look as if they are.

The VRML example given has plenty of specification to help guide development - so it didn't shock me that this was easy to prototype etc.

Where things come tumbling down is where we have Full Stack Salespeople - with all the gear and no idea. They can now produce something that _looks_ like the real deal, but probably isn't sanitising inputs or any of the other "non-functional" aspects of development.

There's going to be work for people in the development/engineering space for the foreseeable future - it's just going to get a bit more niche/specialised and ultimately more expensive IMO.

[rockpile99]

Mixed opinions

Make your mind up Register - it's not that long since you published an opinion article decrying Vibe; https://www.theregister.com/2025/11/24/opinion_column_vibe_coding/ :)

[sabroni]

Re: Mixed opinions

The journalists here actually have different opinions so articles can contradict previous ones. That's a useful attribute that prompts debate and discussion.

Read both the articles and make a decision for yourself.

(also, use an anchor tag and you'll get an actual link in your post: https://www.theregister.com/2025/11/24/opinion_column_vibe_coding/)

[rockpile99]

Re: Mixed opinions

Thanks for the tip about tagging.

My comment was slightly tongue in cheek but when I started consuming news it was through print. When newspapers/magazines put 'opinion' on something then it generally represented the views of the paper (or at least the Editor) - this has stuck with me.

As for Vibe coding, I think it's fine for hobbyists but like all AI not worth killing the planet for.

[Anonymous Coward]

Re: Mixed opinions

I get the reference to Opinion pieces in Newspapers.

I noticed that in the US of A papers it is common to have opinion pieces by multiple people in the same paper.

The cynic in me thinks that this is a way to NOT have an opinion for the Newspaper and that readers will favour the piece they agree with without assigning the view to the Newspaper as a whole. The opinion pieces are always 'Left vs Right' to cover the political landscape of the US of A.

Everything in the US of A is politics and taking a side on ANYTHING is always translated to a political standpoint, to avoid losing readers you publish multiple opinion pieces to 'balance the debate' [Roll eyes] and to take NO actual standpoint as a whole.

There are rare exceptions, usually because the Newpaper owner is actually running the Newpaper as a mouthpiece for a political side they support.

:)

[MiguelC]

Re: Mixed opinions

I'm not really fond of echo chambers, I like reading multiple different opinions in newspapers, and to learn other angles of the same story from people I might disagree with but you also might make good points on the subject

[Anonymous Coward]

Re: Mixed opinions

I am not asking for an echo chamber ... BUT false 'Balanced Articles' are just as offensive !!!

If the Newspapers are posting articles on all sides of an argument it is the same as 'clickbait' on the 'interWebs'.

They believe nothing and are simply trying to get as many readers as possible ... because people will ignore the article they disagree with and believe the newspaper supports their view.

I want a newspaper to stand for something even if I disagree with the stance.

I will read a number of 'left and right' wing newspapers because it gives me a sense of what that group thinks.

I am able to integrate the supposedly opposite views myself.

A newspaper that navigates the political landscape ONLY focused on reader numbers serves no-one and can change its 'tack' at any time.

You cannot rely on anything it publishes as it stands for nothing, it follows the crowd when reporting a story because it does not want to offend the majority of its readership. First to say someone should 'hang' because that is what the majority want and it sells newspapers, last to apologise when they get something wrong because no-one says sorry !!!

Newspapers are NEVER neutral no matter how much they may try ... pretending to be balanced convinces me least of all.

:)

[nijam]

Re: Mixed opinions

> The opinion pieces are always 'Left vs Right' to cover the political landscape of the US of A.

I think you mean "The opinion pieces are always 'Right vs slightly more Right' to cover the political landscape of the US of A."

[Anonymous Coward]

Re: Mixed opinions

<grin> I can see your point !!!

I am using the words that the US voters use and in the sense they understand them !!!

I, personally, think Left & right means something completely different BUT that is because I am not there and can see a wider spectrum of political views.

I am not in the US of A and doubt I ever will go there again.

(Either by choice or because the 'Political Wall' being built would exclude me !!!)

:)

[David Hicklin]

Re: Mixed opinions

> As for Vibe coding, I think it's fine for hobbyists but like all AI not worth killing the planet for.

I think that sums it up, and the big elephant in the room as already mentioned is **maintainability**.

Sure get an AI to write your app but what to you do when a security flaw (or flaws) is found ? Ask the AI to fix it and get yet more code lettuce that now might not work the same and have newer bugs?

[Pascal Monett]

Re: Read both the articles and make a decision for yourself

I've read both articles and I've already made my opinion known : I will believe in Vibe coding when you can ask it for a complete web-hosted ERM and it spits out a fully working copy of Salesforce.

There is nothing in this article that changes my opinion. If you have be an expert developer and you still need to spend two days coaxing this mess into something that ressembles something useful, this is most definitely not "creating code without experience".

Vibe coding is absolutely not warning you about that on its homepage, where it specifically states "making app building more accessible, especially for those with limited programming experience ".

I call bullshit, and this articles proves it.

[HereIAmJH]

Re: Read both the articles and make a decision for yourself

What about when a company decides to do a toolchain shift? Think y2k. How many jobs would have been lost if we had AI to convert COBOL to C++? What about feeding the Linux kernel into AI and tell it to generate it in Rust? What if a company could eliminate all their legacy codebase in a couple months using AI and a small staff of testers? There are hundreds of millions of <insert currency of choice> in potential government work that could be eliminated by AI updating neglected systems.

I don't think it will make app building more accessible. I think it's going to make it automatic and eliminate the need to write code. It's a few years in the future, but not very many years.

[ChoHag]

Re: Read both the articles and make a decision for yourself

So you're saying that as far as programming technologies go, this is the last one you'll ever need?

[Anonymous Coward]

Re: Read both the articles and make a decision for yourself

Why not take the Linux kernel source and ask your favorite, or least favorite if you enjoy a laugh, to write you a new one.

[Anonymous Coward]

Re: Read both the articles and make a decision for yourself or have 'faith' it will work eventually

"I think it's going to make it automatic and eliminate the need to write code. It's a few years in the future, but not very many years."

I think that you are simply hoping it happens ... I think it is going to prove to be another false dawn !!!

Which 'Wish' is more powerful in this battle of 'Faith' and when does the 'faith' translate into reality ???

P.S.

I am still waiting for my Flying car ... it still is not there and is nowhere near becoming a real everyday thing that I can 'drive' !!!

At what point did people who are, usually, very much anchored in reality transition into evangelist who are anchored in 'Faith & Hope' ???

As far as I am concerned 'Faith & Hope' is simply 'grift' painted another more acceptable colour !!!

:)

[teebie]

Re: Read both the articles and make a decision for yourself

How many lives would have been lost if we had AI to convert COBOL to C+

[Roland6]

Re: Read both the articles and make a decision for yourself

>” What if a company could eliminate all their legacy codebase in a couple months using AI and a small staff of testers?”

Legacy code is ALL code being used, including Windows 11. So eliminating legacy is the subject of dreams, not reality…

[Anonymous Coward]

Re: Read both the articles and make a decision for yourself

Trust me, you don't want it to do that... The last thing we need is a salesforce clone.

[PRR]

Re: Mixed opinions

> it's not that long since you published an opinion article decrying

Good (not bankrupt) journalism is a lot about contrasting opinions.

Some of the greatest journalists bickered with themselves. John Dvorak would write LOUD opinions one week and OPPOSITE opinions the next week, or even the SAME week. He spread it around a couple dozen rags, but he was ALL over PCMag. Enraged me so much that I HAD to buy every issue. iPhone bad for a while, then iPhone good. Same for about every O/S and killer-app in the end of the 20th century. Jerry Pournelle tilled this same field, not so blatantly.

[Edward Ashford]

Re: Mixed opinions

To be fair to Jerry, his early articles were full of the sales blurb and high hopes for the thneed (if you haven't read The Tale of the Lifted Lorax I recommend it) he had been gifted, and subsequent articles described the grim reality of trying to make it work in Chaos Manor, including how helpful the supplier had been.

It was a sad day when Byte went digital.

[Anonymous Coward]

once but faster and more chaos

Why does this just sound like the mixed horror bag that was cottage industry of ms access dB to solve everything.... Badly. VB apps that commit all sorts of crimes against coding and logic. Yes users can be empowered, but we've been trying to undo those house of cards messes for decades. Now there techbroa are trying to make it happen after with less controls.

I see a future full of failing apps leaking data and causing disruption at the worst times as the skills to deal with the fallout wither and die.

[AMBxx]

Re: once but faster and more chaos

I'll take your Access & VB and raise you to the horror that was Lotus Notes.

[Pascal Monett]

As a Notes developer for the past 30 years, I have to disagree with you. As far as developing is considered, I have never found a more complete and useful tool. You want something accessible from the web ? The Domino server is all you need. You want a webservice ? Domino server. You want a full-fledged document management system ? Domino server.

With Microsoft, you need at least five different servers, plus their backups (because we all know how well a Microsoft server stands up to the test of time).

Now the user experience is really old at this point in time, I will not argue about that.

But as a developer, it's a dream environment.

[Korev]

When the tool you have is a Domino Server hammer, everything looks like a nail

[Anonymous Coward]

Agree with the dream environment and domino has security built in from the ground up.

Don't think that the user experience is that bad on R14 or higher. to be fair.

From my point of view, even worse are SaaS solutions that take the place of domino apps.

This tends to be for opex reasons and ends up costing considerably more and with severe limitations in capability and functionality.

And the cost of changing / adding something not in the base product will cost you a fortune.

SaaS system often come with their own ticketing system so potentially there is a cost overhead to link the 3rd party's system with the businesses own.

Do that with a lot of apps you could have run off one platform and you end up managing a mess of 3rd party integrations.

[nijam]

Based on my experience, I could add "You want a pile of shit? The Domino server is all you need."

[AMBxx]

The best thing about Lotus Notes is that it's so easy to create an application.

The worst thing about Lotus Notes is that it's so easy to create an application.

Yes, you can create an excellent application if you know what you're doing. However, it's also possible to create an application if you don't know what you're doing.

I still have nightmares about the director who created a timesheet application in Lotus Notes. Needless to say, he had 31 columns to store the hours worked in each day of the month.

[Anonymous Coward]

Re: once but faster and more chaos

and the current cottage industry of Excel spreadsheets to solve everything or end user computing as it's euphemistically known

[sabroni]

It's not vibe coding when you know what your doing

Vibe coding is a non-developer with no development experience driving an LLM to make code.

[Pete Sdev]

"They're now good enough to do things well"

They're really good at deleting hard-drive partitions, for example: https://www.theregister.com/2025/12/01/google_antigravity_wipes_d_drive/

[Anonymous Coward]

Re: "They're now good enough to do things well"

Just not neccessarily the 'thing' that you intended !!!

More and more I am reading promo for 'AI' which seems to be at odds with reality !!!

'AI' has not suddenly got better ... the promotion has just stepped up a factor or two.

If 'Vibe Coding' is so good, why are we no seeing any large corp or industry praising the value of it to them and the real 'positive' impact on their bottom line.

P.S.

Praise from people with a 'stake' in 'AI' selling more does not count !!!

So no 'MS' and their recent '30% of code is produced by 'AI'' !!!

:)

[ChoHag]

Re: "They're now good enough to do things well"

> So no 'MS' and their recent '30% of code is produced by 'AI'' !!!

Oh I'm not sure, this one might actually be true. Microsoft have really outdone themselves in the past year or so.

[Brewster's Angle Grinder]

Re: "They're now good enough to do things well"

And the article links to that story.

[An_Old_Dog]

If a Vibe Coding Tool Wipes Your Drive or Repo

... then, "You're not steering it right."

[b0llchit]

I think I need to find my exploit kit, generate some interesting VRML 1.0 files and send them to you for yourboth our pleasure.

[Dan 55]

Not even OpenAI put its money where its mouth is and vibe coded their browser, they span up another Chromium fork. That tells me all I need to know.

[Jusme]

If this was slashdot...

...I'd tag this article as flamebait :)

4GL

Rapid Application Development

Low Code

Vibe Coding

All wet dreams of MBA-types wanting to use cheap labour to replace expensive skilled workers.

Just like most things "AI", they look good at first sight ("Look at this amazing app little Johnny has built"), but are likely to cause so many issues with security, data integrity, reliability, stability, scalability, maintainability, yada yada... that the lifetime cost will be much more than it would have been if a skilled developer/team was employed to create it in the first place. Of course the manager of the initial development will have trousered his bonus for delivering something on time, and has long-since moved on.

[Doctor Syntax]

Re: If this was slashdot...

I think you're mixing too many different things in there.

As far as the only 4GL I've used (Informix) it was simply a matter of adding abstractions for handling a TUI, relational database (by incorporating SQL) and report generation to what was more or less structured Basic. These were aspects of application development which had come to the fore after the original 3GLs had been developed. It was a very sensible thing to do an provided more productive way of writing applications than writing ESQLC directly*, so much so that when one of the older ESQLC applications needed substantial changes it got rewritten in 4GL.

RAD gets applied as a label to all sorts of things. On the one hand the earlier Informix tool, Perform could be classed as RAD although to do anything particularly advanced needed linking ESQLC into its interpreter. And I've certainly come across an example that had got out of hand and had to be rewritten. OTOH as a fast prototyping tool it was very useful for exploring the essentials of what was needed with a user.

It also gets applied to tools such as Delphi which is fair enough in terms of putting together the GUI - a lot more straightforward than building it with hand written code but is still gong to need the rest of the application to be written in code and in my view its abstraction of relational databases was a good deal clunkier than I4GL.

None of these obviate the need for skilled development - they simply provide abstractions for aspects of application development that were not requirements when the first 3GLs were scoped.

Reducing the skill level needed for development has been a dream almost from the start of computing. COBOL and FORTRAN were devised on the basis that business and scientific users respectively could write their own programs. You just have to consider each attempt on its merits.

* It could compile to an intermediate code to be executed by an interpreter or to ESQLC.

[An_Old_Dog]

Re: If this was slashdot...

COBOL and FORTRAN back then: easier than machine language and assembly language.

COBOL and FORTRAN today, with object-orientation and a giant grab-bag of other features added in: perhaps not easier than the machine languages and assembly languages of back then, but probably easier than the machine languages and assembly languages of today. (Intel's errata for modern x86 chips is longer than then entire MC6800 programming manual. Seriously.)

[Vometia has insomnia. Again.]

Re: If this was slashdot...

I quite liked ESQL/C back in the day; I did some serious development with it, as well as some useful side-projects but with the latter my manager insisted they be rewritten in 4GL because it was The Future™, in spite of my cow-orker (who was the only 4GL programmer I've worked with before or since) telling him that in her opinion one of the main requirements was impossible to accomplish using Informix forms. AFAIR the project was canned as a result, which was quite an unusual gaffe for someone who in all other respects was the best guy I've ever worked for.

[may_i]

Re: If this was slashdot...

Lovely to see all the old Informix programmers putting their hands up.

I wrote acres of ESQL/C code back in the day. Loved it! There was some 4GL involved as well, but I try to forget about that.

[Anonymous Coward]

Re: If this was slashdot...

Been there, done that, didn't break anything. We still hadn't made it 'intelligent' before I left, but it could check that what you'd entered made sense*.

* - as far as we could tell.

[nijam]

Re: If this was slashdot...

> All wet dreams of MBA-types wanting to use cheap labour to replace expensive skilled workers.

... when, ironically, vibe coding would be only just about capable of replacing MBA-types.

[Filippo]

I like reading sci-fi novels, and one thing that's always true of futuristic computers is that they are apparently very easy to hack. Whether it's an astromech droid or a teen in a hoodie, if the heroes need to get through a computer system, there's someone in the party that can do it in seconds, or minutes if there's some need for drama.

Each time, I wonder: this is set centuries if not millennia in the future, the story imagines amazing advances in all fields of human knowledge... except computer security, which appears to have gone backwards? Why?

Well, now I know.

[Anonymous Coward]

Ask and you shall be answered !!!

Answers:

1. Because Security is hard to do well and computers are also hard to 'do well'.

2. Security is ALWAYS added on at the end because building it in is hard (See 1.) and it slows down producing/testing the code.

3. If it is hard to do and costs money to do as well someone will always take the shortcut and HOPE they get away with it !!!

4. Bacause no one looks back in history to learn any lessons ... apparently !!!

5. Because Sci-Fi is based on real-life, at some level, therefore see (1. & 2. & 3. & 4.)

6. Because without some flaw somewhere the Sci-Fi story would be very short and sad as the main actors are caught and killed.

:)

[xyz]

I read a Google "vibe coding" gush-agrada thing last night...

In the end some "associate" bloke had "vibe coded" an html front end using Google's latest and greatest, that looked like it had crawled out of a Contoso example from 1999. No back end, thank God although Google reckons that's soon. The whole thing looked like a non functioning joke nailed into GitHub and lurking on Google Cloud.

I think we're safe.

[DrewPH]

Re: I read a Google "vibe coding" gush-agrada thing last night...

Is there any chance you could provide an English translation of this comment?

[doublelayer]

Re: I read a Google "vibe coding" gush-agrada thing last night...

Some definitions and translations you may find useful:

Contoso example: A fake company frequently used by Microsoft to demonstrate how a service might be set up using a specific Microsoft product. "Example" because it wasn't a full implementation, but a usage demonstration of a specific component. Microsoft assumed people reading the examples would decide which components to use, so they were showing how to use each one, not telling you which ones you needed for any given task. Therefore:

"looked like it had crawled out of a Contoso example": Had the form of a solution without the content and was missing important elements that needed to be added.

"No back end": The code that was produced only shows the interface. You need the back end code to do any tasks. Given their assessment of the quality of the front end (part you see), they weren't very confident that it would have managed the rest.

Does that help?

[DrewPH]

Re: I read a Google "vibe coding" gush-agrada thing last night...

Contoso yes (never heard of it in my 25+ year dev career) but was really hoping (or is that fearing) to know what gush-agrada means.

[doublelayer]

Re: I read a Google "vibe coding" gush-agrada thing last night...

Both parts of that are words for substantial amounts of praise. Gush is a common term in English, whereas agrada sounds like it was either taken from one of a few romance languages where it's a word meaning the same thing or as part of one of a few English words that come from the same Latin root such as gratify, congratulate, or ingratiate. So I think you could translate that as "I read an intensely self-congratulatory thing from Google". I'm quite certain that's the intent from the gush part, so if they meant something else by the "agrada", they can correct this.

[Bebu sa Ware]

"rewrite expectations about how IT will operate before the end of this decade."

You mean it can get worse ?

[ecofeco]

Re: "rewrite expectations about how IT will operate before the end of this decade."

Not can. Will.

[stiine]

Re: "rewrite expectations about how IT will operate before the end of this decade."

Not will. Is.

[Anonymous Coward]

Re: "rewrite expectations about how IT will operate before the end of this decade."

Always.

[BJC]

It's going to be interesting

Currently a human developer will consider edge cases and exceptions during the initial design. That avoids things slipping between the cracks. An AI assistant isn't going to take that approach and so it's going to take a lot more effort to interpret the code to find those issues, or lots more testing. For example, how do you know that the code doesn't include functionality that handles an input value of 55% differently than any value, just because some training data had that quirk?

The difference of doing development as an occupation, rather than a sideline, is that code is written not just to be functional but to avoid issues and aid long-term support/maintenance.

As others noted above, we've been through all this before. It's easy to throw together a form that accepts a numeric value and there's a button to calculate the result. Fine. Now, what about when the value is out of range - is that handled? What happens when the user enters a non-numeric value? If the calculation needs to be changed, is the calculation clear, or filled with magic numbers? Can the calculation be changed for another, or is it all tangled in the UI code? Oh, and has there been revision control to understand why it's working for Fred but not Wilma?

That said, how many here have had a manager that looks at the rough code and observes that the project is basically done and just needs a software dev to "release it"?

[Doctor Syntax]

Re: It's going to be interesting

"Currently a human developer will consider edge cases and exceptions during the initial design."

Only if the PM makes provision for doing it properly. I think we can all think of examples where that doesn't happen. Even "initial design" appears to have been superfluous in some cases.

[doublelayer]

Re: It's going to be interesting

The more people there are who understand this, the more it happens, and yet even if managers and PMs don't understand it, there is a certain amount of coverage that programmers will do all on their own. I know that, if I don't make this work with certain error cases, they're going to happen and I'm going to have to fix it and I'm going to be yelled at for why it broke in the first place, so I might as well make it work now. If the PM who didn't understand this complains about me doing that, it's probably easier for me to explain, possibly dishonestly*, that this is necessary to keep it from blowing up than to try to do that after they have an error they are blaming on me.

* I fortunately have mostly had managers who understood what I was talking about so I've rarely had to. I will admit that I have, at least once, lied to people that something I was fixing was done to prevent serious consequences even though, in practice, we might have avoided the edge case altogether and it wouldn't have been terrible if we hit it, just because I didn't want to live in a world where the code was guaranteed to fail if we did. In that situation, we did end up hitting that edge case (from logged data), so I count it as having saved at least a day and some user complaints, but that was not guaranteed.

[HereIAmJH]

Re: It's going to be interesting

That said, how many here have had a manager that looks at the rough code and observes that the project is basically done and just needs a software dev to "release it"?

I have never had a manager qualified to look at any code. Yet when the date arrives that they have given to upper management, they say "release it".

[Peter Prof Fox]

Hurrah!

No more £1.8 billion for the national ID card project. That should be ready to trial next week then.

Seriously though this sounds like a viable tool for prototyping. I've got specs and proof-of-concepts lying around but whizzy to have a demo people could have in their pocket.

[Brewster's Angle Grinder]

"...run-once disposable software."

In my experiments I've found them "better" at writing new modules than handling poorly documented, legacy code bases which don't work quite like as they predict. So I can see software becoming more disposable and it being easier to scrap the existing thing and write a new thing. Whether that's a module or a whole app.

[Larry D]

While it's still useful and I use it there are limitations at the mo.

A person, way above my paygrade as CTO at top 10 global bank, said it was not currently scaleable because of the effort in checking the results. That rings true from my experience.

"it" i.e. AI cannot currently fully understand context. So (say) generated SQL joins can be silly. You can, I presume, get around this with a sufficiently detailed system prompt describing he data model. However, that can be a lot of work to write. And some data models are, for no good reason, nuts.

I have always written nice modular code that is well documented. Mainly because I'm an old fart who can't remember what/why I coded what I did. The idea of trying to debug this AI code is something else. I have found the AI generated doco looks excellent but is sometime plain wrong.

So I wonder how maintainable vibe code is is . Or is that just another AI prompt + learning?

I've been coding since the 1980s and I will stick with vibe coding. But be sensible.

F

[HereIAmJH]

"it" i.e. AI cannot currently fully understand context. So (say) generated SQL joins can be silly.

There are already plenty of apps using ORMs. ORMs don't generate the best SQL, but often what they do create is good enough. Sometimes is comes down to paying a 6 figure salary for a developer vs buying a larger server.

[breakfast]

Finally we can get rid of the boring activity of writing code

At last we get to focus on the really fun part: Quality Assurance.

That's what we all dreamt of when we did our comp sci degrees, right? The hope that one day we would have the opportunity to test bad code written by a machine that doesn't understand anything.

[MrAptronym]

Re: Finally we can get rid of the boring activity of writing code

I am so glad we can all get slightly more work done by simply becoming assistants to a machine and giving up the act of creation to all be editors!

Of course, the company would need to get more work out of you to pay for the $300 worth of credits, and you certainly aren't going to work shorter hours. Now that you are just QA though, perhaps you don't need that fancy developer salary?

[David Hicklin]

Re: Finally we can get rid of the boring activity of writing code

Dream on ; The customer does all the QA these days

[Mike 137]

"good enough to do things well"

"it's already “good enough” to display all the examples from my 1995 book on VRML, and other bits of content I found online"

reminds me of the numerous occasions I've reported to some web dev that their offering crashes, and got the reply "well, it works for me"

[Someone Else]

Re: "good enough to do things well"

QA: This PoS doesn't work right. It crashes when I enter '42' into this field.

Dev: It works on my machine....

QA: Yeah? Well we're not shipping your machine!

[Jason Bloomberg]

Distant horizons

It's now possible to forecast a time - before the end of the decade - when these tools have been sufficiently "softened" to allow pretty much anyone within an organization to rapidly develop an app for a specific use case.

What is the basis for this assertion; that vibe coding will go from requiring domain knowledge to 'anyone can do it' within five years?

What is there beyond wishful thinking?

[O'Reg Inalsin]

Re: Distant horizons

Exactly! The author segues from

I wrote a few VRML 1.0 browsers thirty years ago, so I came to this with deep domain knowledge. If I'd asked Antigravity to create a fault-tolerant database that would preserve transaction history across power failures, I'd have made a muddle of things because I don't know enough in that domain to steer the machinery. That means domain experts in software engineering and operations aren't going anywhere - but they will be getting a lot more productive.

immediately into

It's now possible to forecast a time - before the end of the decade - when these tools have been sufficiently "softened" to allow pretty much anyone within an organization to rapidly develop an app for a specific use case.

without any evidence at all from the entire excellent narration of his adventure. At least he made it obvious.

"Nothing to see here, please disperse" Naked Gun [https://www.youtube.com/watch?v=rSjK2Oqrgic]

[Anonymous Coward]

Same old starting point

... that I'm wrestling with on a non AI related project right now. The requirements have to be good enough and the customer has to understand them well enough to say that the product is doing what's expected.

If you're going to rely on an AI tool to generate a system for you that meets your requirements + complies with relevant legislation + any other relevant governance etc. then I would think your prompt is going to be a mighty one.

Now the pace this stuff is moving at I suspect it wont take long before many of these hurdles are overcome but ultimately it's not the AI that gets hauled before the judge if your system leaks sensitive data.

[David Hicklin]

Re: Same old starting point

> your prompt is going to be a mighty one.

It might even start to resemble a working system...

[Roland6]

Re: Same old starting point

I see no one has picked up on the line:

“ Antigravity had done just enough work to render very simple VRML 1.0 files but had left vast portions of the specification unimplemented.”

Ie. It had ignored vast amounts of the carefully crafted prompt.

Something that doesn’t bode well for testing, repeatability and rework.

[Anonymous Coward]

Example of vibe coding in action

My father-in-law used vibe coding to quickly cook up a simple webpage with multiplication problems for my son to practice on. It looks pretty good.

Of course, some of the answers are wrong...

[User McUser]

Re: Example of vibe coding in action

Is this post supposed to be ironic?

"AI is fine, it failed to correctly teach my child basic math skills" is not exactly a ringing endorsement...

And could your father-in-law not be bothered to write up some simple math problems on their own? No? They pawned it off on the computer? I presume they love their grandchild, but I guess not enough to put any real effort into it?

[Anonymous Coward]

Re: Example of vibe coding in action

The site was intended to autogenerate as many problems as desired, rather than be a one-time-use thing. And I said it **LOOKS** pretty good, not that it IS good. (Which it isn't, since it gets the answers wrong.) So yes, it was ironic - it looks good, but it's actually terrible. Vibe coding in action.

[Nathan 6]

Vibe Coding -- Great for Hobbiest

From my own experience, the new crop of LLMs do a pretty good job of generating mostly working code for hobby projects you wouldn't otherwise bother doing, given the time involved. This is especially useful for Arduino/ESP32 projects. I recently spent time working on an audio visualizing app using the M5Stack Cardputer Adv and HTML/JS app. As far as production use, I would not trust the output code as yet until someone with knowledge does code review and extensive testing. They seem to give you code that 90% functional, but just end up in endless loops when ask to provide the remaining 10% functionality or fix something. Checkout the project on GitHub (https://github.com/ns96/CardputerMicTalk)

[Anonymous Coward]

Re: Vibe Coding -- Great for Hobbiest

*hobbyist

[harrys]

hyper individualism rules ok

1. excellent for the short term and ego of the vibe developer

2. terrible for future generations of dumbed down developers running around in nappies at the mo

but hey they wont know how dumb they have become, so probably ok :)

[pacohope]

LLMs are frequently "better than nothing" and not better than anything else

One thing nobody seems to point out is that even when LLMs look their best, it's almost always in tackling whitespace. That is, building something from scratch that has never existed before. You don't see a lot of people saying "I've been maintaining project XYZ for 15 years. Now I'm using an LLM to handle pull requests and add features to my 15-year old project faster than ever." The reason is because it doesn't work that way. It's not that nobody is trying. It's that the tech can't do that. We have to wait for all of these novices—who have never spent a decade maintaining something important—to learn a bunch of lessons that became common knowledge across the industry decades ago. The world runs on legacy code. Everything technological thing we count on in our daily lives is delivered by OLD code.

[MrAptronym]

Re: LLMs are frequently "better than nothing" and not better than anything else

I pray for the people who will be stuck maintaining today's bespoke vibe-coded projects years from now.

[mattaw2001]

They're now good enough ... if you take the time to learn how to s̶t̶e̶e̶r̶ t̶h̶e̶m̶ program [n/t]

They're now good enough to do things well, if you take the time to learn how to s̶t̶e̶e̶r̶ t̶h̶e̶m̶ program.

[Groo The Wanderer - A Canuck]

Which AI companies paid for this "opinion piece" to be published? This kind of shilling of proven-broken technology is way beneath The Reg.

[Omnipresent]

What they are telling us

is that you will have to pay a subscription fee for the privilege of using a computer that you pay exorbitant amounts of money for, so you can be harassed, lied to, manipulated; and trolled by super criminals and dictators from now on. Even on linux. No where is safe, no one is safe, because you were too eager to do what the tech bros wanted. You could not put down the computer, and just like everything else in this God forsaken existence we now live in, it's entirely your own fault. You could not help yourselves.

[AdamWill]

Suggested headline alteration

"Vibe coding will deliver a problematic proliferation of personalized software"

The problem I'm seeing with this, so far, is that, yes, people who don't really know what they're doing can vibe code a thing that mostly works!

So...they do.

They *all* do.

Now instead of having three or four tools for doing the thing, each maintained by someone who vaguely knows what they're doing and in which effort to do it well can be centralized, there are 50 tools to do it, all built by people who don't really know what they're doing and have no concept of how to maintain software in the long term, each with their own bugs and (no doubt) security issues.

Can you imagine having to sort this mess out in a few years when companies start realizing there are tens of thousands of vibe coded scripts doing stuff with their data and they have no conception of exactly how many, who is using them to do what, or what's wrong with each of them?

yikes.

[David Hicklin]

Re: Suggested headline alteration

> having to sort this mess out in a few years

Sure, no problem...just create another one using an AI and now there will be 51 tools to do it...

[Ribblethrop]

That is fascinating. However details are missing. Please provide your ‘prompts’.

This would be great information, for prosperity, before AI is able to guess what we want a priori.

[Always Right Mostly]

I just used vibe coding and it was awsome! Had a "Hello Wordl" app in seconds.

[Ken Shabby]

Genius you’re hired.

Now write some control software for that Nuclear power plant we are building.

[Guido Esperanto]

A couple of points to add:

1: The author had already compartor solutions with which to test the coding against - for expected outputs and potential pitfalls.

When building something "new" where are your guard rails, which brings me to point 2

2: You say you are an experienced programmer and I'll accept that at face value. But you're not the target audience for this. Those with "limited" programming knowledge.

I mean "limited" is one hell of an undefined marketing attribute.

Reading a chapter of a book takes you from no knowledge to limited knowledge in about 10 minutes.

[may_i]

LLMs write shit code

I've just had a project dumped on me because the person who wrote it has retired.

The guy who retired was a systems integrator, not a developer. He used CoPilot extensively to help him write the project.

The code is nothing short of a disaster! He and CoPilot have no idea how to properly handle exceptions and unexpected errors. The resulting system is so fragile that the author spent every day massaging the system manually to correct all the problems that the code sometimes creates when it fails. Program logs were published to a combination of a file and the Event Log making tracing what the system does a total nightmare. Concepts like validating input data before trying to process it are completely absent from the project. There are methods which are many hundreds of lines long. In one place, the code just continues after receiving an exception while trying to access a database. Quite what it does at that point is unknown.

LLMs are not fit for purpose when it comes to writing anything more complex than a one-off, throwaway utility. Regardless, the companies who are pumping up this multi trillion dollar bubble are telling senior people (who don't understand the technicalities) that they can fire all their developers and jive code everything.

Dark times lie ahead.

[Moldskred]

Why "will", not "is"?

So why aren't we seeing a deluge of new software and small software startups?

[MacGuffin]

Redundant Department of Redundancy

"integrated IDE"

So...

An integrated integrated development environment.

I'm shocked, SHOCKED! /s

[doublelayer]

Re: Redundant Department of Redundancy

If we're going to be pedantic, and you started it, then those integrateds don't mean the same thing. The original integrated meant that the development tools were all together, integrated with one another. The second one means that those tools, still integrated with each other, are now integrated with a code guesser too.

That's neither small nor obvious. The old version involved copying and pasting code into and out of models, and it was a lot worse if you wanted to play roulette with your code generation. How to keep all of that code in the context window was tricky. Now the software is theoretically able to manage it for you. I doubt it does it well, but that's something that was quite a bit harder with standalone models. Other coding models managed it with IDE plugins which had similar goals, although depending on how much work Antigravity did, it could easily exceed those in feature set, and if I had to guess based on the amount of destruction it can cause, I think that it is likely they have managed a larger integration than their competition.

[Crellin]

April Fools

Is this an early April Fools - or will next weeks opinion piece be from the same commentator asking why Agentic AI has deleted his repo?

[The Onymous Coward]

I was a big fan... for about half a day

I do some development as part of my current not-actually-a-dev role. Much of what I build is supporting tooling for far more complex pieces of software, for example, an automated testing framework for a HTML5 UI built using Puppeteer.

One enterprising young chap took my code and fed it into GPT to add a new feature. He proudly showed me his (/its?) work, and I have to admit, I was impressed. He added a few more features over the course of the day, then called me to his desk for help with an intermittent issue where the test run would bomb out with an exception (not always the same exception!)

Over the course of several iterations, the code had become largely unrecognisable. It took me all of two minutes to decide that I wasn't going to support whatever had been built by GPT.

If that is what it can do to a simple Puppeteer tool in a day, I dread to think what a proper software project would look like after a year in the wild.