vibe-coding, your thoughts?

Vibe coding, just like all AI generated coding, is done by people who don't really love coding, and in most cases people who don't truly understand what the AI is producing.

Reading code might sound easy at the surface but is actually the hardest part of coding, as it requires engaging with someone else's train of thought. It's easy to be convinced about the logical soundness of something if we aren't deeply involved in researching that thing, just like it's easy to be swayed by populist propaganda if we take everything at face value, because there are a number of caveats that, without involvement, we can simply not be aware of, so the first line of defense when it comes to quality and safety in software are the developers who write the code themselves, and that line of defense simply doesn't exist if the code is AI-generated. Furthermore, anyone who isn't skilled enough to implement a solution themselves is also totally incompetent to review other people's implementations, and that includes AI-generated code.

Even if AI was really good at coding and truly raised the ceiling, as is apparently the case with theoretical math, there would still be fundamental problems like token costs currently being heavily subsidized, the impact of data centers and their cooling needs in nature, the the economic inequalities, and the general dumbing down of human intelligence. I mean what would our purpose be in a world full of machines individually more intelligent than us? Do we really want that future? I certainly don't, I value humans and their creativity way too much to let go that easily, so I'll never let anyone including AI prevent me from learning and gaining experience. Each time we outsource the solutions to our problems is a lost opportunity to learn.

At the end of the day, I just want a tool that actually works. Call it vibecoding or whatever else you want—AI is everywhere now, and I get why people freak out about code breaking. But honestly? Give me software that solves a real problem, and I couldn't care less who built it.

Without getting into the debate, here's something useful from Devin Prater on Mastodon (@pixelate@tweesecake.social).

Blind Vibe Coding

A screen-reader-friendly directory of resources for blind and low vision people who use a screen reader and build apps with help from AI.

This directory lists English-language resources for blind and low vision people who use a screen reader and want to build apps with help from AI, sometimes called “vibe coding.” It covers guides, tools, courses, communities, podcasts, articles, and research, free and paid, for Windows, Mac, iOS, Android, Linux, and the web. Every resource explicitly mentions blindness, screen readers, or accessibility. Every resource was published or updated in 2024 or later, and every link was checked on 1 October 2026.

https://jamalmazrui.github.io/BlindVibeCoding/

Agree with maldalin and others, enough said.
Its worth noting that so what if someone vibe codes? Obviously you have to try and do your best. In order to avoid mess ups. But the opposite is also true, plenty of codersAbsolutely suck at it. Plenty of coders probably right terrible code. Like with anything else in life, there are people who are incompetent at their job, lazy, they don’t put the effort, it’s not a matter of, oh if you don’t bother learning reading code, etc, or Coding You Love it. no, I asure you, there's plenty of people who just code because, it’s a job, and nothing else. They don’t necessarily love it. They might actually hate it for all one knows, and in any case if they love it or not doesn't matter, since it pays the bills? You get the point. now. Does having coding knowledge, etc Help? Yeah, no shit, it makes things easier. But saying that oh vibecoding isn't loving coding, not how life works, but I digress. In the end, one can only do their best at what they're doing, that's about it. Not to mention the obvious... There is way more to coding than just reading and writing code. You could be the best programmer / coder in the world, but if you don’t have the proper ideas of what you want to code, how to make it i.e as far as logic, ideas to bring the architecture to life so to speak, what do you need to do to make it happen, you could be the greatest coder in the world, and you’re application, software, what have you, could still turn out to be absolute garbage. Just look at an example of all accessibility problems of say, meta? All the accessibility regressions? or insert name of software that has terrible ui or ux?point made. That software was coded more than likely by people. Point is, its not just coding, coding is a branch of a much, much bigger tree, if you will. Anyway, those are my thoughts.
As far as the child labor stuff? First? Completely out of topic. Second? Absolutely nothing to do With the current discussion. Thirdly? You cannot do anything about it. Suppose you stop buying from a company that you know uses child labor. Guess what? There are likely dozens more companies who use child laber. Whether you choose to not buy from one, yeah, good luck, you’re likely buying things from 10 other companies that also use child labor. You cannot do squat to stop it, or change it. Also? Funny, well ironic that you mentioned child labor. Why? Guess which company has used child laber? in the passed? Weather they still do or don't doesn't matter; they have done so in the passed.
Apple.
https://humantraffickingsearch.org/resource/new-forced-labor-and-child-labor-charges-for-tech-giant-apple/
That is just one link. I can pull up more, and of more companies besides. What I’m going towards, is that was rich. Coming from you, someone who owns multipleapple devices? using your logic? You bought a Apple device? Congratulations, you're supporting Child labor...
I myself of course don’t go for that philosophy, I think it’s bullshit, and it’s way too black and white, but enough said.

Okay, let's get away from child labor and back to vibecoding, please... Thanks 😊

Hi all,
I vibe code a lot. I actually released one thing, a static HTML page with a list of resources as part of the mailing list I run. I tested it extensively before releasing it, and I know a bit about accessible web development which helps, so I was able to tell Codex from ChatGPT how I wanted the site to be designed. It took me a few hours to test and debug but it was worth it. I still add things to it from time to time. Another significant thing I tried to do: I uploaded a pull request to a developer on Github, because according to some people on Mastodon it had some security holes. And when I took steps to reproduce said security issues they were there. So I got the code from Github, put it through Codex, and asked it to suggest ways of improving the security, and I got it to look at another NVDA add-on from a developer I know who implements good security for his add-ons. Said developer can actually code by hand, but he can code from scratch so he knows what he's talking about. In short, I did a bit of research into Windows DPAPI, told the add-on to implement things, and then I tested it to make sure security issues had gone away. The issue was the API keys were being saved in NVDA's ini.txt file and in some log files too. But the fix I vibe coded meant that Windows DPAPI was being used to hide API keys, and they weren't being saved in NVDA .ini files. I tested this and it did actually work. So, I had a security fix, but my PR got rejected. The developer gave me a lot of feedback as to why, which actually helped. He didn't merge the PR I submitted. But I later implemented the improvements he suggested and now use my fork as opposed to the original. One of the reasons it got rejected was because Codex hadn't first implemented the security everywhere like I'd thought. Keys were still being exposed in NVDA's log file under certain circumstances. It got to the stage that I hadn't remembered what I'd tessted and hadn't tested, but either way, that slipped through. So the disadvantage of vibe coding is it can miss some security implementation. It might suggest security fixes, but it doesn't necessarily implement them across the board, it only partially implements them. This is where vibe coding really falls short. It's the sort of thing where a developer has to know all the possible scenarios a vibe coded security implementation might miss. I've tried vibe coding a few NVDA add-ons, but never finished them, because if you want something without bugs you really need to test for hours. I vibe coded myself an iOS app for listening to the radio with an interface and features I want, but I certainly won't be paying for an Apple developer account to release it, and it's far too buggy to be released anyway. It's got bugs that annoy even me. Vibe coding isn't a problem at all, provided you test your software extensively before releasing or giving your code to others to work on. The reason some developers seem to hate vibe coding is because they're given vibe coded stuff by colleagues to work on, but because they've just stuck it through an LLM with no testing.

I don't see an issue with this as long as the person who's doing it tests their stuff first. More power to ya if you know programming as you'll be able to direct the AI in where you'd like to go more so than not. Use AI as a tool, I always say, not a lazy get out of jail free card. I've been working on some python apps for personal use, not ready at all and worked on a few php scripts that has a ton of useful things but I always test and review my code and if I need to either make changes myself or have AI do it. I've seen some really nice apps out there, clipman comes to mind and if I know the guy as well as I think I do, he don't code much at all. So in short, yes, you can code pretty decently as long as you know prompting and some at least, basic coding. Garbage in, garbage out. The laziness comes with just assuming that AI is gunna get it right 100 percent of the time and you don't bother testing at all and releasing a product that don't work half the time. That's the problem I have with all this.

I'll just come right out and say it. I think vibe coding is absolutely revolutionary. Hands down, one of the biggest game changers we've seen in the world of programming, and I'm genuinely excited about where it's taking us.

And I'm saying that as someone who used to program for a living, long before AI was even part of the conversation.

Back when I still had my vision, programming was a big part of my career. I loved it. There was something incredibly satisfying about taking an idea, figuring out how to make it work, and watching it come to life. Man, if we'd had something like today's AI coding tools back then, I would've been like a kid in a candy store! I can only imagine what we could've accomplished.

But life took a different turn for me.

When I lost my vision, I eventually moved away from programming and more into networking. Not because blind people can't program. Far from it! There are some incredibly talented blind programmers out there, and I have nothing but respect for them.

But speaking strictly for myself, programming became much slower, more difficult, and honestly, a lot more frustrating than it used to be. Something that had once come so naturally to me suddenly required so much more time and effort. Eventually, I found myself drifting away from something I'd always loved.

And that bothered me more than I probably realized at the time.

Over the years, I occasionally dabbled in different programming languages, but it never quite felt the same. That spark just wasn't there like it used to be.

Fast-forward a couple of decades.

Along comes AI. And then vibe coding.

Holy cow. Talk about turning my world upside down!

For the first time in years, I feel like I've gotten a piece of my old self back. Something I thought I'd pretty much left behind is suddenly part of my life again, and I can't begin to tell you how much that means to me.

And here's what I find especially exciting.

Just look at what's happened in our own blind and visually impaired community over the past year. We're seeing more and more people experimenting with vibe coding, building apps, creating software, and bringing their own ideas to life.

Think about that for a moment.

People who may never have considered programming before, or who found traditional programming too intimidating, time-consuming, or frustrating, are now creating things they might only have dreamed about a few years ago.

How in the world is that a bad thing?

To me, that's absolutely incredible!

We're talking about people being able to create tools that solve real problems in their lives. Apps that fill accessibility gaps. Software that does something they've wished existed for years. Ideas that might never have gotten off the ground otherwise.

And perhaps best of all, we're giving people a chance to discover something new about themselves. That feeling of accomplishment when something you've imagined actually works is pretty darn amazing.

Now, let's talk about something else that's been floating around these discussions.

This whole idea that you're not a "real programmer" or "real coder" because you use AI or vibe coding.

Oh, give me a break.

I've seen people suggesting that vibe coding makes you lazy, that it's somehow bad for your brain, or that you're not really accomplishing anything because AI is writing the code.

I'm sorry, but I think that's a bunch of crap.

Let me tell you something from my own experience.

Back when I was what some people would apparently call a "real programmer," I worked with all kinds of programmers. Some were absolutely brilliant. I'm talking about those naturally gifted people who could look at a problem and come up with solutions that would leave me scratching my head in amazement.

And no, I definitely wasn't one of those naturally gifted geniuses! I worked hard at what I did, and I was pretty darn good at figuring things out, but I knew plenty of programmers who could run circles around me.

I also worked with programmers who were average, programmers who cut corners, programmers who were downright lazy, and programmers who wrote some truly awful code.

Believe me, crappy software existed long before anyone ever heard the words "vibe coding."

I spent plenty of time taking over somebody else's project, fixing broken code, cleaning up messes, or trying to figure out what in the world the previous programmer had been thinking.

So let's not pretend traditional programming was some magical world where everyone wrote beautiful, efficient, bug-free software.

It wasn't. Not even close.

And you know what ultimately paid my paycheck?

It wasn't whether somebody thought my code was elegant enough. It wasn't whether I used the fanciest programming techniques or whether another programmer would've done something differently.

It was delivering something that worked and did what the customer needed it to do.

Period.

Now, does that mean code quality doesn't matter? Of course not. It absolutely does. Security matters. Reliability matters. Accessibility matters. Testing matters. Being able to maintain and improve your software matters.

But those things matter regardless of whether a human typed every line of code or AI helped create it.

That's really my point.

The tools you use don't automatically determine the quality of what you create. The thought, care, effort, and attention you put into the finished product make a tremendous difference.

And that brings me to something I've been working on.

For about five months now, I've been developing an AppleVis iOS app, which I'm hoping to get into the hands of private beta testers in the near future.

And yes, I'll say it proudly.

It's 100% vibe coded.

There. I said it!

But don't mistake that for five months of sitting around telling AI to magically make me an app while I kick back and watch television.

Far from it.

There has been an enormous amount of thought, planning, designing, experimenting, testing, fixing, rethinking, and refining involved in this project.

I've spent countless hours thinking about how things should work, how the experience should feel, how VoiceOver users will interact with different features, and how to make the app something people will genuinely enjoy using.

And I've learned a tremendous amount along the way.

One thing I've discovered is that the more you work with AI coding tools, the better you become at using them. You learn how to explain what you want, how to break down complicated problems, how to recognize when something isn't right, and how to keep improving the results.

AI doesn't magically eliminate the need to think.

If anything, this project has given me more opportunities to think creatively and solve problems than I've had in years.

And I absolutely love that.

Will my app have bugs? You bet it will!

Show me an app that doesn't.

Will there be things that need improving? Absolutely. That's part of developing software, whether you're vibe coding or writing everything by hand.

I've seen some absolutely incredible apps created with AI assistance, and I've seen some pretty terrible ones, too.

But guess what? I can say exactly the same thing about traditionally developed apps sitting in the App Store right now.

The difference often comes down to how much effort someone is willing to put into making their creation better.

If you're going to vibe code, then put your heart into it. Don't just accept the first thing AI spits out and call it finished. Test it. Challenge it. Break it. Fix it. Ask questions. Learn from the mistakes. Keep polishing until you've made something you're genuinely proud of.

And please, especially in our community, don't forget accessibility. Just because AI helped build an app doesn't mean it's automatically accessible. That still takes thought, effort, and plenty of testing.

I'd be willing to bet that if you put a well-designed, thoroughly tested vibe-coded app next to a well-designed, traditionally programmed app, most everyday users wouldn't be able to tell you which was which just by using them.

And really, why should they have to?

They want something that works, meets their needs, and gives them a good experience.

Now, if somebody wants to tell me I'm lazy or that I'm not a real programmer anymore because I'm using AI, that's perfectly fine.

They're entitled to their opinion.

And if the fact that my AppleVis app is vibe coded bothers someone so much that they don't want to download it, that's fine, too.

I really don't give a shit.

I'm not building it to win some imaginary contest over who deserves to call themselves a programmer.

I'm building it because I love doing it, because I believe it can be useful, and because it's given me back something I thought I'd lost a long time ago.

And that last part is probably the most important thing I want to share.

Some of you may have noticed that I've been pretty quiet on the podcasting front for much of the past six months.

It's been a really rough year for my family, particularly with my wife's health. I won't go into all the details here, but it's been one of those stretches in life that can really wear you down, both mentally and physically.

And I'll admit, I haven't always been in a very good place.

There were days when I had far too much time to sit around and think about everything that was going wrong. It's amazing how quickly your mind can wander into some pretty dark places when you're dealing with things you can't control.

Then I started working on this AppleVis app.

And something changed.

Suddenly, I had something to look forward to each day. Something to get excited about. Problems to solve. Ideas to explore. Features to build. Little victories to celebrate when something finally worked the way I'd imagined.

Instead of spending so much time dwelling on the negative things happening around me, I had something positive to pour my energy into.

And I can't tell you what a difference that's made.

It's helped me mentally. It's given me a sense of purpose and accomplishment again. It's helped me stay engaged and more positive during a time when my family has really needed that from me.

Most of all, it's reminded me of the person I used to be when programming was such a big part of my life.

I honestly haven't felt this excited about creating something in a very long time.

And to think that something called "vibe coding" could have that kind of impact on my life!

Who would've thought?

That's why I have such a hard time understanding the negativity surrounding it.

When I see another blind person excited because they've just created their first app, built a little utility, or finally figured out how to make something they've wanted for years, I don't look at them and think, "Well, you're not a real programmer."

I think, "Hell yeah! Look what you just accomplished!"

That's something worth celebrating.

We spend so much time in the blind community talking about barriers, inaccessible technology, and things we can't do because somebody didn't consider our needs.

Now we've got technology that's opening doors and giving people new ways to create things for themselves.

And somehow we're supposed to feel bad about using it?

Not me.

And here's another thing. Some of my old programming buddies from back in the day are just as excited about AI-assisted programming as I am. These are people who spent years writing code the traditional way, and they see the potential, too.

They're using AI to work faster, explore ideas, solve problems, and become more productive.

They're not sitting around mourning the good old days when everything had to be done the hard way.

They're embracing what comes next.

That's how technology has always moved forward. The tools change. The possibilities grow. And people find new ways to accomplish things.

AI-assisted programming isn't perfect. There are real concerns about security, quality, overreliance on AI, and understanding what your software is actually doing. Those are fair conversations to have, and we absolutely should have them.

But there's a big difference between having a thoughtful discussion about the limitations of a tool and putting people down for using it.

You can acknowledge the problems without dismissing the possibilities.

And you certainly don't have to discourage someone who's just discovered a new passion.

I remember somebody once telling me a joke.

What do you call a programming student who graduates with a D?

A programmer!

Ha!

Now, obviously, there's a lot more to being a skilled professional programmer than simply getting a passing grade. But I've always liked the point behind that joke.

Not everyone starts with the same abilities, takes the same path, or develops the same level of skill.

Some people are naturally gifted. Some struggle. Some learn quickly. Some take years. Some write beautiful code. Others need a little more help getting there.

And now some people use AI to help bring their ideas to life.

So what?

There's room for all of them.

And for those of you reading this who are curious about vibe coding but haven't taken the plunge yet, here's what I'd say.

Go for it!

Seriously. Give it a try.

You don't need to start by building the next great iOS app. Start with something small. Maybe a little utility you've always wished existed. Something that makes your daily routine easier. Something fun. Something silly. It doesn't matter.

Just start creating.

You might discover that it's not for you, and that's perfectly okay.

Or you might discover something you absolutely love.

You might even find, as I did, that it brings a little piece of yourself back that you didn't realize how much you missed.

And to those of you who are already vibe coding, please keep doing what you're doing.

Don't let the negative comments or the people arguing about who's a "real coder" take away the enjoyment you're getting out of it.

Take constructive criticism seriously. Learn from people who know more than you do. Listen when somebody points out a genuine problem with your software. That's how you get better.

But don't waste your energy trying to win over people who have already decided that what you're doing doesn't count.

Sometimes the best response is simply to keep creating.

As somebody once told me, arguing with people who are only looking for a fight just gives them more of what they want.

So let them argue.

Meanwhile, you keep building.

Keep experimenting. Keep learning. Keep improving. And most importantly, have some fun with it!

Because at the end of the day, life is too darn short to let somebody else's definition of a programmer determine whether you're allowed to enjoy creating something.

I don't know exactly where AI and programming will take us over the next five or ten years, but I do know that we're witnessing something pretty remarkable.

And for someone like me, who spent years thinking that a chapter of my life had pretty much closed, it's been an incredible experience to watch that door swing open again.

Vibe coding hasn't just helped me create an app.

It's helped me rediscover something I love, given me a much-needed positive outlet during a difficult time, and reminded me that losing one way of doing something doesn't necessarily mean losing the opportunity to do it forever.

Sometimes a new tool comes along and changes what's possible.

And sometimes, if you're lucky, it changes a little more than that.

For me, it certainly has.

So whether you're a seasoned programmer, a complete beginner, someone who's blind and always wanted to try coding, or just somebody with an idea you've been carrying around for years, don't let anyone convince you that you shouldn't give it a shot.

Create something. Make mistakes. Learn. Have fun. And be proud of what you've accomplished.

I'll be over here happily vibe coding away on my AppleVis app, enjoying every minute of being able to do something I thought I'd lost.

And if that makes me something other than a "real programmer" in somebody else's eyes?

Well, I can certainly live with that.

Because for the first time in a very long time, I feel like I've got that old programming spark back.

And that, my friends, is worth more to me than any label somebody wants to put on it.

sure, it is hard work to be a programmer, and I commend the people who are. now, Vibe coding solves real issues. take, for example, TurnOver, which I am beta testing. that developer fully is transparent with where the code comes from. I submit feedback, and for build 16, he mentions me by name in the release notes. as long as you've got human beta testers, , and not AI, they can help iron out the bugs. can AI do that in future? who knows, this is still early technology. Vibe coding makes the world of programming, accessible . isn't ' that ? all that matters? I learned some Python myself, and would like to learn another programming language. . the code I wrote, has been obsolete for quite some time. I wrote it using Python 2, which modern Macs do not run. it would be interesting to look at that old code, and upgrade it, for lack of a better term, for Python. three. Maybe. I'll write some more, just to make sure my skills are still up to snuff. now, some cannot code, and Vibe coding is great for those people. for those who enjoy it, power on! for those who don't, Vibe coding is an amazing way forward. Weather it's right or wrong is subjective, but I am all for it, as long as it is tested by humans before release. AVM, another vibe coded program, unfortunately seems to be abandonware, as V. 2.3 was. a preview release, whose last update was in August. correct me if I am wrong. that solves another problem for us, who are fortunate to have Macs with Apple Silicon. is this project still actively being developed? the issue with Narrator not working, saying exiting Narrator, for me, is documented on Github, with the status of open. to code that by hand, for lack of a better term, would be an enormous undertaking. nI commend the developer for their testing, but Vibe coded software, traditionally teeters out, as coders end up running out of money to pay the outrages fees, for something you can learn yourself. it'll be interesting to see, is AVM and TurnOver stay around. so it's a double edge sword.

Lately, it seems like there are an awful lot of blind vibecoders out there. Some things they've made are pretty neat, but it's github I loathe with every fiber of my being. It's obviously a me problem, but I find it all but impossible to download stuff sometimes. Part of me would love to jump on the vibecoding bandwagon, but I know absolutely nothing about programming, and don't feel as though it's anything I could learn at this stage of my life. So, for now, if I hear about something that sounds fun or useful, I will try to check it out, but more often than not, because of the whole github thing, I just read about stuff and then go on with my day.

Hi All,
Oh Thomas I really love your podcasts. I hope your wife gets well soon. And Jonathan thanks for the info on Clipman, I might check it out.

Vibe coding helps me to break away from in app purchases and having to wait for someone else to strike my fancy on something. I rarely host things on Lovable, and mostly entertain myself with games in html format. I don't need to answer to others, and not everything has to earn a spot on the app store either. Vibe coding in html format also helps me keep things truly free. I can share it with those who ask or those I know, people can modify their own copies or trash it for all I care. It doesn't get more free than that.

I find this such a difficult topic - I have such conflicting opinions about Ai in general.

As mentioned before, I write code (mostly Python) for a living and have done for many years now - only the last few without sight. During that time I have picked up specific ways of doing things, and I agonise over code quality and trying to make sure I have the correct abstractions, or that things are named properly and understandable. In Python there is PEP8 which sets out how to format files - two line breaks between functions, things named a particular way. For God's sake don't put more than one blank line at the end of a file. All this kind of thing. And I've always been about making sure that everything is as good as I can make it. I want to be proud not just of what something does but how it does it.

A few years ago I started having to use a screen reader and had to re-learn so much of my job and my life. Now the other side of that, I can coed but I find it as much frustrating as rewarding.

And now today we have vibe coding that threatens to take away the one thing I know how to do, that basically renders the last however many years seemingly redundant, that my skills don't matter any more. And this combined with my sight loss have made it difficult for me to reconcile my place in the world.

I also have to wrestle with the idea that maybe all my pedantic ways are actually not really useful or contributing to anything. That maybe no other human is ever going to need to read my code so many things just don't really matter any more. I've not quite lost my need for control in my work but I can see it happening.

So that's all to say that I cannot have an opinion on vibe coding that is entirely unbiased.

At the beginning of the year, I was given a ChatGPT licence where I work. I use BitBucket Cloud and for a year or so they had completely broken my ability to do pull requests through the web interface. It had become such a struggle and was really getting me down and they refused to do anything about it. The ticket was resolved and marked fixed a good year after I raised it but it still remains pretty unusable.

So I asked ChatGPT to write me a native Mac app that could connect to BitBucket, let me open up a pull request, see all the changes in a diff, make comments, read and reply to comments and so on. I kept extending it, adding in the ability to view pipelines and monitor them. And now I almost never touch the web site and use my app. The UI isn't going to suddenly change when I need it the most and I can easily tweak and fix things as I go. I wrote a companion app that helped me diff locally in git and then ended up with a load of other features around that. I wrote an app so I don't have to fight with Jira, or Spotify. I gave up on the idea of Jetbrains ever fixing the accessibility of DataGrip and wrote my own. I wrote a game to help me practice braille on my Orbit Reader. Now I have something like typing of the dead in braille, and it encourages me to play about with it more which is great.

All of these apps have made my work and beyond more enjoyable. I've been able to make them all do exactly what I want and it is really empowering. And I don't know Swift and have never looked at a single line of the generated code. I realise this is reckless but it hasn't come back to bite me so far. I'm not sure I would be confident releasing any of it but it's made my life better.

I think the power of vibe coding for many people isn't that they can write something to top the app charts - they just write something that does exactly what they want and who cares about anyone else. This is a new type of programming - before we were always the slave of the developer and had to fit our ways of working around them. This flips it on its head - now we are in charge as the consumer and the code does what we want.

It seems that anyone who has tried vibe coding, particularly someone who has not really coded much before, has suddenly felt a massive rush of adrenaline and excitement about what it makes possible. And I think this is both good and bad.

Steven from Double Tap talked a lot about the item chooser he made for NVDA. This is something he wanted that didn't exist - it was a single thing that fitted a specific purpose. I've not tried it but this sounds like the perfect use case for this.

On the other end of the scale, you have apps where you can tell someone has got over-excited. Quill always comes to mind - it is a text editor that has a billion features but just isn't very good at editing text. Bear in mind it is still beta, so it's unfair to criticise it too much but it does illustrate an easy tendency of the vibe coder. It is much, much easier and more fun to add in a big sparkly brand new feature than it is to fix a bug. And again rather than fix a bug you also end up with an AI slop web website with a million words saying almost nothing.

My heart always sinks when a beta or version 1 app comes along with so many features because you know that the vibe coder probably isn't acting very responsibly. The overall quality cannot be that good if a product is so big from day one.

As a consumer of apps I am finding it quite hard to reason about all the vibe coded apps that are being released. For starters, there are just so many of them. Before a new navigation app would have got me interested, now I just roll my eyes.

The thing is, when you manually code an app you will spend a lot of time on it. It becomes a labour of love. You keep the number of features down to make sure they work. When you finally release the app, you will have gone through the initial excitement and into long, hard graft in trying to get everything to work. You will have invested a lot of yourself in it. So I think by that nature, you have made a commitment to the app which has already been proven by the fact it got released.

But these days you hear someone saying "I spent so long writing this first version of the app - almost a whole week - so can you guys all test it because I can't really be bothered". Actually, the testing is the graft, so it's just being offloaded onto the end user.

I am not saying that writing an app must be slow, tedious and difficult. But if something is this easy, how do I know as a consumer that you are in it for the long haul? Your intentions are likely very good, but no doubt you are already coding your next big thing and the first app that I am apparently testing for you is now no longer new and shiny but a bit boring.

As a screen reader user, learning a new app requires time and effort on my part. Getting good with it usually needs a lot of that. Now I don't want to waste my time just because someone got over-excited over the course of a long weekend, and now I'm using something that is already gathering dust.

I'This isn't a problem with vibe coding as such, it's more a problem of making a difficult, often tedious thing, too quick and too easy. What are you going to do when the novelty is gone? Are vibe coders really going to be knuckling down and tackling the tricky awkward little bugs. Because an app like Quill already does everything in the world, so all that's left is bug fixes. (Well, maybe it just needs a built-in podcast player or two, who knows)

I think when something is difficult it earns respect and the responsibility is obvious. When something is thrown together and easy, it just doesn't have that.

That isn't to say that all vibe coded apps will be like this, but it does mean that I must sift through hundreds of pretty crappy apps to find anything good, and even then I'm not going to know if it is worth my commitment.

This is really the Spider Man principal - with great power comes great responsibility. Sure, write a fun little diversion and shove it on github or whatever for people to play with. But I think if you are going to come up with a fancy web site, put it up on the app store and start hyping it, then know that doing this does come with responsibility even if the app is free. It might not cost money to use but it does cost time.

In terms of actual code quality, I suspect that vibe code is much better than I would like. I want my code to be better, but it probably isn't in a meaningful way. I use AI to write some code at work, but only on some things and I always carefully review. It does have a tendency to over-complicate, and often you will ask "why did you do that?" and it will reply "oh yeah I didn't really need to do that". But on the whole it's actually doing the job. And the code in my Mac apps - well who knows what that looks like but it's working for me.

How much code quality matters depends on what the software is doing and the context. Something that runs millions of times a second in the cloud might need to be more efficient than a little game running on your phone for example. A lot of criticism of vibe coding seems unable to accept that not all software is the same.

One final thought (for now) and it's something I've had when writing a lot of my apps. We are suddenly finding that an app is something that everyone in the world can do without any effort, easy peasy. And suddenly we're all excited by that. But I wonder how long the concept of an app is even going to last, because it feels to me that AI is even making that obsolete to some degree.

For example, take my Jira app. The other day I was trying to find a ticket and just couldn't. I asked ChatGPT - can you find a ticket, I think it was about this sort of thing. And it thought about it and up it came. There is actually so much you can do just by talking.

One thing AI is good at is making sense of data. So a typical app might have me trawl through lists of this and that, an each is unrelated so I need to mentally track keep track of things as I work through it. Then maybe I need to go to the database or look in this other place, or refer back to an email or whatever. Really hard to do in an app, very easy for an AI.

With AI I could receive a Jira notification with a new bug. The AI could see which account in our test environment this relates to, go find out whatever information it needs from whatever data source, merge it all together and then present it to me however I like. This could mean I don't have to go through several pages in our back office and mentally map things around, or hunt down through the databases or whatever. . Catering for all these kinds of permutations in an app just isn't feasible, but it's easy for AI.

Or maybe I know someone sent me something but I don't know if it's Mail, iMessage, WhatsApp, Teams etc. I'll just ask AI what it is and now I'm not using an app at all I'm just getting the information.

I'm not saying apps are dead but I am saying that I think over time we will become less dependent on them and more dependent on our AI overlords who will keep our brains pickled in a jar should we ever need to refer back to them.
I wish vibe coding didn't exist - I want to keep my job. But

Sure, I'm coming at this from the perspective of an author, but I've got thoughts.
Firstly, I'd just like to say that Thomas, I hope things get better for you and your family, and that I'm glad you've found some joy among all the difficult stuff.
Second, I'm not a fan of generative AI. That's because with all these people out here using AI to code, there comes millions of apps. That floods the market, and therefore it gets even harder to sell your app if you've built it from the ground up. That minimizes all the testing, coding, and whatever else goes into building an app. I'm a writer, not a coder, so I've got no clue.
Then we've got books and music. All that AI generated stuff again floods the market, making it nearly impossible for the human-made stuff to get a foot in the door. For creativity, nobody will ever convince me that there's less heart and soul in something that wasn't built by a human.
And lastly, to get fully back on topic, we've got people out here "coding", who have no idea what to do if AI breaks their stuff. Sure, you could just prompt it to do something and maybe fix it, but then you've potentially got an even bigger mess on your hands that could have been solved just by knowing lines of code.
What it all really comes down to is creativity, a flooded market, and the ability to make your work stand out. And with all the AI clones of everything, people just throwing whatever into an AI and call that creating...yeah. I've got concerns. Sure, this is an unpopular opinion, and I'm glad for all those for whom this gives you joy and makes you feel as if you've gained a new perspective, but take your internet connection away, and you've got nothing. If you're gonna do this kind of stuff...put in some work. Code some lines. At least learn. Don't just let the AI do it all for you. Not saying any of you do this, but nobody can argue that there aren't millions of people throwing words at AI and just running with whatever it spits out.

The amount of accessible apps, and alternative screen readers we've seen coming on to the market recently, is partly due to vibe coding. Some of the screen readers on Non-Apple side have blown existing screen readers out of the water.
I think it's okay if you are developing something for yourself, but if you are putting it up on the internet for everyone else to download and use, it's got to go through some thorough review process. Few months back, I saw a storage usage warning on my laptop, and I was totally confused about it. Turns out, one of the AI description addon for NVDA was so poorly designed that everytime NVDA started, it downloaded a small offline AI model, again and again. My appdata folder was full of it. Originally I had downloaded, played with it for a while, and kind of forgotten about it. Any ways, removing said addon, and clearing out that specific folder helped me claim back that storage.
I've also seen lot of vibe coded apps which work for others, but refuse to work for my machine. And mine isn't that really special.