As if we already didn’t have a problem of bad app pollution and had to sieve through all the garbage to get to sth reasonable.
Step 1: Vibe code an app.
Step 2: It’s buggy as shit.
Step 3: You have no idea how to fix it because you don’t actually know what you’re doing.
Step 4: Your app gets terrible reviews and nobody uses it.
Step 5: Release your next vibe-coded app, because failure doesn’t mean anything anymore. If developing an app is free and easy, it doesn’t matter if it’s a failure since you really had nothing invested in the first place.
I’m going to say something extremely controversial on something that I, as a ex-writer, lost the battle against a long long time ago.
Vibe-coded apps are like self-published ebooks. If you are a crap writer, and your book doesn’t sell, there’s no consequence to it. If nobody buys it, or it gets bad reviews, you don’t really need to care, because it’s entirely free to just keep throwing your crap out there no matter how bad it might be. You can literally bang your firsts on a keyboard for a couple of hours, pay a few bucks to upload it to Amazon and call youreself Stephen fucking King. It doesn’t matter if it doesn’t sell because it literally cost you pennies.
A legitimate publisher would tell you to go fuck yourself, of course. But self-publishing is like flunking out of med school and decided to set up a portable operating table on the hospital sidewalk anyway because “How dare some gatekeeper tell you that you’re not good enough.”
Vibe-Coders are of the exact same ilk. Gatekeepers exist for a reason, in both publishing and app development. If your work is crap, the response is to GET BETTER, not say “Fuck it, I’m doing it anyway”
Sorry…rant over. I get a little passionate about this particular subject.
LLMs are destroying the “small time author” ebook world, too. Unfortunately. Now the bar is typing prompts into ChatGPT for story paragraphs, and you can publish that junk on Amazon en masse.
Also…
LLMs can be a useful ufor coders who know what they’re doing, and use is strictly as a utility.
And for my personal writing, using local LLMs as a glorified thesaurus, hacking them into completion syntax and viewing their raw logprobs, really helps me over some mental stumbling blocks in my personal writing.
It’s mind boggling that text prediction models, which were always supposed to be dumb utilities akin to a calculator or dictionary, have been twisted into what they’re presented as, today. It’s insane.
It’s like everything that can be turned into a grift, becomes a grift, and people are lapping it up.
I totally agree. Unfortunately my country doesn’t really have real publishers anymore. They only publish books from politicians, actors or books from other countries. They pretend to make an effort with small competitions but they only publish short stories and generally you can trace the winners to famous people anyway…
If you try to publish there will only be fake publishers that want your money. Literature is basically dead. I’m not a disillusioned writer by the way, but I know one (i can’twrite shit so I know how hard it is to actually write a book). It’s so sad.
At least developing is not going away because companies will always need some software and sw created via vibe coding sucks so hard it makes me laugh when people think it’s somewhat usable.
No users. Pffft, we’ll just make more AI users! That definitely won’t kill the earth. /s
I remember when people came to me after they saw ‘the social network’ to pitch me their ideas. I’d be doing all the work but I’d get like 30% or something, because the had the idea, which in their minds was the only thing that counted. The ideas were usually ‘facebook, but with <insert one useless feature>’.
These kind of people can now finally make their own apps, and the result is exactly as shit as you’d expect. Even worse due to garbage AI code.
They are all hitting the same hard truth: making the app is kind of the easy part
In their very half-hearted defense, some total weirdo did become one of the richest humans ever by saying “Myspace, but more superficial and with Harvard’s PR machine backing it.” Several subhuman ghouls become insanely wealthy by saying “banking, but online”. There was also “Walmart, but online”, “record store but on the phone”, “all this open source software but it’s mine now”, and several other giga-genius inno-venters.
Absolutely, ruthless opportunism definitely works. Especially for people with money. Paid-for expertise definitely produces a lot better apps than if the same people tried to do the design themselves.
Tech CEOs think they are the only creative people on the planet, and that their creativity is worth all the world’s money.
Hard to say for sure, but it looks more like they’re towards the bottom on the creativity scale.
Reminds me of those people claiming they’re “10x more productive” and “doing the work of entire teams”. With so much increased productivity, I would assume these results would be plainly visible through rapidly improving featureful applications, but instead we see declining quality in established software (e.g. windows 11) and a whole lot of garbage nobody wants to deal with like slop PR spam and these app releases.
They’ll refuse to admit the obvious decline in quality at the major tech firms too.
It’s horrible. I have a client whose whole setup is based on Google drive and the APIs shit the bed constantly. It’s only getting worse they are so careless about these products that they might as well be legacy.
Trying to coordinate initial code for a new project - the guys had done a lot “because ai”.
…… but none of the tests actually verify anything …… and half of them still manage to fail
The place I work for has a hardcore vibe coder and today I found out none of the e2e tests pass because there is no db setup but it’s okay it doesn’t make the ci red as ignore_failed_tests is set to true
AI generated test cases are exactly the fight I’m about to have to go through. It’s gonna be such a metric pain in the ass
This is a narrow area where ai can help, but like anything else, someone needs to go over it. I’ve used it successfully here, but I evaluate the results and iterate it correctly its mistakes
Last time I worked with this guy he had generated more test cases than there were on the entire product, yet his module still managed less coverage. Lots of useless test cases, lots of test cases that don’t verify anything, lots of duplication. Most importantly it’s generating lots of code that now has to be run on every build and needs to be maintained forever
People see this and think “AI is just generating a bunch of useless slop.”
What they need to be thinking about is how much of our environment was destroyed to produce this useless slop.
Like, imagine all of those bullshit cheap plastic nic nacs that flood our oceans and landfills, times a million.
Not saying you’re not correct but I keep hearing, what I take to be at least slight overstatements, of how much environmental impact is caused by agentic ai but is there some good source where I can red up on exactly how much? While ”Plastic nic nacs that flood our oceans and landfills, times a million” is surely a good ballpark I’d like to narrow it down a bit
Exactly this. It is not that AI generates meaningless slop, it also destroys the environment and increases costs to populations near data centers…
Dont forget the decimated value of the houses nearby.
From a stupidly high price like 1.3million to just horrible loss values like 1.25 million.
They’re not building these datacenters near million dollar homes. It’s the average person who suffers.
Average homes cost that much now.
Not in the vast majority of the country, no. They do not.
Exactly my point. It’s sort parallel rather than pro or against.
It’s a self solving problem. As soon as people pay the actual cost of AI, usage will drop automatically. It’s just because investors are willing to burn money inflating the bubble is the resaon AI is being used as much as it is.
I’m using Opencode and the free Deepseek in it pretty much daily, how is this destroying the environment and increasing costs to other people?
It feels like there’s a push on the internet against AI because it’s a largely western product.
Look up datacenter water usage and utility pricing - not to mention so many studies showing using AI makes you dumb and unable to think critically.
Because you are using exponentially more energy than you did before AI. And all for something that at best could have been made without AI, and at worst is completely frivolous and useless.
And this at a point in human history where we desperately need to cut down on energy usage.
I knew we were truly fucked the moment I realized that even non-AI content on the internet is affected negatively by AI. Take this comment I’m typing out right now. I’m not using AI for it, but the content of what I’m writing is based on experiences I have had with AI - nothing is purely human made anymore.
At one point I had an issue with a tool and searched for it on the web and found someone who looked like a real person describing a similar-ish issue. They mentioned trying a setting but it didn’t work for them. I thought “that setting seems to be what I need”. I searched for that setting on the web and also in the codebase of that tool - zero matches. It was made up. I didn’t use any AI myself, but I still wasted time with an AI hallucination.
Now every time I find something on the web I have to do additional searches to see how many other sources corroborate the information I get from any result. And eventually this too will be pointless.
It was made up. I didn’t use any AI myself, but I still wasted time with an AI hallucination.
And then it gets used as training data.
More like, imagine all the aquaphors drained dry.
Not to be that guy, but aquifers.
Inference is about as cheap as playing a videogame.
There’s certainly a lot of externalities to AI and enviro impact is absolutely one of them but this isn’t a great point for it.
I play vidya gaem ⏩ I have a pleasant time
AI does things ⏩ …?
As someone that goes back to the early 90s in my career, and who lived through the whole dot com boom… My perspective remains unchanged from decades ago.
Without diminishing the skill it takes to make a quality app, the fundamental truth that I could never get across to people is that an idea, some underlying tech infrastructure and an app are the smallest part of making a business work. The business part is the hard part.
These tools have made it easier than ever for people to run off half cocked and make the same mistake that’s been made for 4 decades. And they appeal to the worst sensibility, the incorrect assumption that if you build it they will come.
If its anything like the Play Store, the vast majority of these new slop apps are just shady copies of existing apps designed to milk their popularity for AD money, or worse ship malware onto your device.
Play store is hot garbage, god I fucking hate their shitty ui filled with ads, and all the reviews are fakes.
F droid!!
Of all the trends in software development (and here, I use the term very loosely), shovelware certainly is one of them.
It’s certainly the most recent
Shovelware has been a thing since the Atari 2600.
It’s because they never found out about my calculator app, which was totally amazing, and not basic at all.
What it has spurned, as usual with tech, are talentless hacks trying to make easy money for nothing.
Weird how reviews peaked and then dropped, would assume entire business are being build on fake ai reviews
Maybe Apple is pretty good at filtering out fake reviews?
Maybe reviews increased due to all the new apps and then fell off as people began to identify slop apps.
the absolute cratering of app reviews
I like the latent reduction in “Apps with significant usage”. I will assume it’s due to the abandonment of apps after developer adoption of vibe code or AI as primary dev tool and subsequent reductions in performance, utility, or ergonomics. Or alternatively/additonally that good apps just get drowned in the slop noise.
A few weeks ago, I downloaded a puzzle app.
Upon opening it, it told me that, by the way, it does not collect any data. Also, here’s a “Privacy Center” in the app to show me how they’re not collecting data. Unsurprisingly, the privacy center did not display anything.
But what was worse, is that the puzzles were either far too easy or just entirely broken.
The pinnacle of ingenuity was when a puzzle asked me to compare four shapes and find the one which is mirrored, but then showed me symmetrical shapes.Trying this app was entirely a waste of my time.
Which is kind of the fundamental problem I see here. There’s so many apps for all kinds of purposes out there, it’s just not useful to create Yet Another Calender App. It’s a waste of time for me to try anything but the most popular app, or well, whatever is on F-Droid.
In theory, there can be hyper-specific niches that aren’t covered by existing apps, but that’s also damn near impossible to explain to potential users, so probably still don’t end up with many users.
