A couple weeks ago, I saw a cool video on YouTube. Hugh Howey, the guy who wrote the Silo books (and is executive producer on the Apple TV adaptation) had made a writing application for writers, and everything about it looked awesome, especially the price tag. Not only was it free, but it was open source as well. I installed it, and I shared it to a writing comm.
I expected to get at least one thanks for spreading the good word, but instead people told me I was posting AI slop. I had no idea! It IS open source, but I didn’t think to look on GitHub. So, sure enough, I took a look, and the authors were hughhowey, and Claude, Anthropoic’s AI chat bot.
I asked Howey for a comment on it on the YouTube video, and he actually gave me a pretty good answer. He said he made the tool he wanted to use to write books, and he figured he’d share it. Also, he said he had been looking for the perfect tool, what he wanted, and he felt it didn’t exist. He also implied he was trying to pay people to write his program, but no one would (at least, at what he was offering). I thought, that’s fair. He made it for his personal use, and he released it for free, and posted the code so that anyone can see it.
I decided against using the program because I don’t want anyone to be able to say that AI helped me write. Although I will say, to the app’s credit, as a Mac user here, anything I write in (current choice is Ellipsus, a privacy and anti-AI focused web-based writing tool; freemium; but also iWork Pages, LibreOffice Writer, TextExit (Mac’s version of Notepad), whatever — Apple “Intelligence” is going to suggest stuff as I write. Happens in Firefox, too; and Safari. Howey’s tool is able to mute that. It also disables the system spell check. Basically he makes it so you can just write, without your OS looking over your shoulder. (There’s a toggle to turn this on and off, for editing.) His reasoning is that if you have a character or town that isn’t in the dictionary, you shouldn’t be pestered by red squiggly lines.
This isn’t the first time I’ve been fooled by vibe-coded AI slop. About a year ago, or so, I downloaded a cool iPhone app where I could enter in something that happened frequently (or infrequently), and log every time it happens. Some call these habit trackers, but I don’t use them to motivate me to keep the habit; I use them to look back. Like when I finished a book, or a movie, or a game. Or when I went to a restaurant, or visited another city or something like that. Also some health/home stuff. So I can show my wife the receipts, as it were.
The app was lacking a few features I wanted (like Dark Mode!), so I wrote the author… after paying $5 for his app. He wrote back and said he didn’t know anything about coding, he was out of AI tokens, and he wasn’t going to pay more until more people bought his app. Like any reasonable person, he was trying to come out ahead, and apparently he wasn’t there (or ahead “enough”). Again I made the decision to drop the app, and I found another one that does the same stuff but better.
The first case, it’s kind of on me because it’s not that hard to look at a GitHub and see Claude as a contributor. With iOS apps, AFAIK you can’t do that. With Android apps, I’m pretty sure you can open an APK in a zip program and look at the Java code. (I’m not a programmer but I know a thing or two.) But back when I was big into modding Android phones, I heard they were going to start encrypting APKs so people couldn’t just steal the code or whatever. I do not know if that ever came to pass. I never learned anything from unpacking an APK anyway.
Is there another way to tell?
Maybe there should be a repository for AI-generated/vibe-coded apps, but it would be like playing whack-a-mole and it would be hard to keep up or maintain it. I suppose it would be possible to ask an AI chatbot if it can detect AI involvement, but that kind of defeats the purpose.
Got a report on this one that it might be veiled advertisement for Howey’s app. I don’t think that’s the case, and I tend to want pretty solid evidence of something before taking actions. So, yeah. I don’t think any mod action is warranted in this case. Just wanted to offer a little transparency on this.
Okay, thanks. I didn’t name it in the OP (or at least, I tried to avoid doing so); however, I did name it in a reply.
I’m a fan of the show and I read the books, but I haven’t read anything else by him. The only exchange we ever had was when I questioned his use of AI.
His app is free. I gain nothing by promoting it. Also, I don’t even use it. I used it for a couple days, until it was brought to my attention that it was made with Claude. Then I uninstalled it.
You can offer a little transparency. I will offer complete transparency. (Not disparaging you, just having some fun with words.) I have nothing to hide here.
Yeah, no worries. I think the report was mostly baseless. I think I’m glad that the report was made just so that the mods here could make a determination, so I don’t think the reporter was wrong to report in the first place, but neither do I think you were.
The dude used AI to write this I feel like too… Just getting that vibe
The answers you’ve already gotten are the best. Commit patterns. By far.
But the code itself has a distinct shape: impeccable application of pattern. Essentially perfect.
Which then is juxtaposed by bizzare, compounding, and amplifying design choices, as if some invisible rule or assumption was invented and became a self imposed constraint to be worked around… with every iteration pointing at the previous as justification for it’s correctness for applying it the next.
Neither in isolation is weird, it’s the marriage of the two. It’d be like watching a pro driver making millisecond perfect RPM matching shifts and then struggle to parallel park. Like… It doesn’t make sense. It doesn’t appear in nature.
It’s normal to see a gap for human devs, although it’s typically inverted (better evidence of justification rational design choices, worse application of patterns). But it’s the MAGNITUDE of that gap (and ALWAYS favoring application of pattern) that gives it away.
Sufficiently supervised processes raise the level of design coherence to roughly that of the operator. So I’m sure I could be bamboozled in those cases. Not sure exactly sure how everyone draws the line for what constitutes “vibe-coding”.
Those are almost scarier though, because the level of QA or user testing would sit at non existent I’m sure.
F-Droid (the FOSS software repo for Android) has been getting more and more AI slop too. Just yesterday, I was looking for a SSH/terminal app, and top/recent results have Claude as author.
Sometimes difficult to tell, but what I notice:
- Sloppy description with emoji-lists
- New app with huge amount of functionality
- Very frequent updates with meaningless descriptions
- Individual elements look well made, but everything does not seem to fit together or make sense.
- Pretty icon (marginal, yet…)
I recommend ConnectBot for SSH.
AGENTS.md and Claude as contributors
TBH I know ConnectBot predates it, and I did not look further into it, but it’s still sketchy to have it in an app that I expect to handle secure connections.
Also commits with more than a thousand changes, and a tree-style list of the files in the README (absolutely useless but AIs love it).
regarding terminal/ssh on android… isn’t termux the best?
That’s what I went with after all :)
Hey, I was looking for SSH client/terminal app the other day as well, but I still haven’t decided what to go with. Is termux really the best? Isn’t it cumbersome to use on a mobile device?
I have found other apps (Conduit, Haven Terminal) that may be better, but I don’t really trust them on the first glance. They are too new and could have been made with AI. Not sure if anyone knows anything about them.
Random markdown files and comment blocks all over the place.
I’ve been in the professional software space for 35 years and one of the hard rules is documentation and code must be kept separate. Each has it’s own place, goals and infrastructure. Often people have access to documentation, even technical in depth documentation, that should not have access to the code. Documentation is often translated in larger multi-national teams and used for all sorts of non-code related tasks throughout the organisation. AI is trained on a lot of tutorials where it’s normal to have huge blocks of documentation in the code and kept as markdown files in a project. It’s mostly meant to be descriptive to others and as an example, a convenient way of keeping everything together for people to learn from.
Now with AI slop coding you see this pattern being applied to all coding with huge amounts of comments, entire blocks of documentation and markdown files in the code. Where this, at least in my experience, wasn’t the case in professional environments in the past. Sure a comment here and there to set people on the right path, but the inclination to write comments is often treated as code smell.
See I tend to over comment as a hobbiest that writes their own code, so that kinda bums me out.
Same here. Even at my job we all comment our code because we spend more time in it and need a quick way to search the code for parts doing specific tasks. I did the same for a class and got a 0 for using AI. Makes me sad too
I think you’re underestimating the amount of comments we’re talking about. I routinely see AI generate files where more than half the lines are comments! And the content of those comments are often inanely detailed and focused on inconsequential stuff, often referring to earlier iterations (“we’re no longer doing X because Y“) when those iterations were never even commited.
His reasoning is that if you have a character or town that isn’t in the dictionary, you shouldn’t be pestered by red squiggly lines.
His reasoning is kinda wonky. Adding custom words to autocorrect makes it more likely that I spell new words consistently because they don’t have red squiggly lines after I add them.
Hey that’s the town I used to live in before I was concieved!
Ah yes, Dwigt.
(In the American The Office, the boss, Michael Scott, writes a screenplay called Threat Level Midnight and has his employees read the script. There’s a doofus, I forget his name, but his name is given as “Dwigt” at one point, and the assigned employee, Dwight, quickly realises that the character was originally called Dwight, and Scott simply find-and-replaced every “Dwight” to the name of the doofus character, but at one point he made a typo, and so that was not caught.)
But yes, that’s what I was thinking as well. “I’ll just add it to the dictionary.”
I use this https://slopscore.ava.pet/
With this browser extension https://codeberg.org/ashkitten/slopscan-webext
As a starting point
A simple function that has clear problems.
I saw an application form online where you couldnt select your country (for your phone number.) But only on the desktop version! On the mobil version it worked! Another one that cleared all inputs when pressing “shift+Q”.
usually the logo is ai, the description is ai, and the creator has like five hundred other apps.
That wouldn’t help me here — AFAIK, it’s his only app. In the case of the writing app anyway.
Quantity is absolutely the tell with YouTube slop channels. You’re really making a whole new song every day? Obvious slop. But in that case, trying to find it on Apple Music is a good tell. Spotify allows it. I thought AM did too but a lot of AI slop channels don’t deal with AM for whatever reason.
And it was created 2+ months ago and has had like 1,000 commits all by one person early on but none of the commits are recent.
The biggest tell for me is when the code makes ridiculous use of commented code. AI loves to comment out everything that it does
I knew a junior who did not speak English yet his comments were at the same level as Shakespeare. I told him “We need to talk…”
return 1; //returneth 1Almost. Every time “he” wrote a comment, it was 5 lines of the most perfect English that I have ever laid my eyes upon.
I like comments, but not fake bullshit.
Many LLM-generated projects are new and only a few months old. I would avoid such projects (regardless whether they declare ai usage or not). Most die quickly and end up unsupported.
For older and larger projects, the cadence and size of submissions should give a valuable hint. For any project that is larger than, say, 10,000 lines of code, a contributor will rarely submit more than 100 new lines of code per day, and very rarely more than 1,000 lines of changed code. Anything exceeding that is a strong hint that primarily LLMs are used.
Some more subtle tells I’ve noticed from Github repositories is if all files were added on the same date (especially when there’s no further revisions) and if the version number starts at 1.0 or 1.0.0. With real open source development they tend to work through revisions of alpha and eventually beta builds, and some will skip “1.0” releases entirely maybe due to influence from the Linux kernel’s versioning being whatever Linus feels like. Hope this helps in any way :)
If you still have a Microsoft/GitHub account for some reason, you could block the user “@claude”. Whenever you go to a GitHub page, it will display a banner saying that Claude contributed. BUT you cannot block @copilot (intentionally), the “show” button doesn’t actually work for “assisted” changes AND the banner will still be there if the change was reverted. So it’s only really a little clue.
Anyways, you are gonna love GNU nano, lol. It’s a text editor. It edits text. Spell checking isn’t even an option. You press a key combo (which I successfully forgot and am shy to admit) to turn on/off text wrapping. There are a lot of other features if you take time to find them, but you are safe to completely ignore them.
Honorable mentions
- mg - also just a text editor, slightly different keybinds modeled after users of a different text editor who now regret using it. Unfortunately doesn’t speak Unicode.
- micro - imagine nano but with Windows style keybinds and a few dozen color schemes. You can copy paste not just within the editor but also within the OS, if you fuck with the settings a bit. Note that if you decide to open (what it decides is) code with it and it has very long lines, it’s gonna spin CPU trying to add pretty colors to it even after you exit, it’s an old bug.
- If you install “plan9 from userspace” or one of the few dozen variants, it’s gonna come with a few graphical text editors. Very specific text editors that make extensive use of a 3 button mouse and single letter keybinds/commands. Light mode only, but it’s the less evil light mode. Some people around Lemmy who aren’t as dumb as me and actually learned one claim they are really good.
Github only shows stuff like claude if you give the bot access to your github repo with write priveleges. Which is insane, but it is definitely a clue that the owner of the repo isnt writing much of his software.
Pretty sure (source: trust me bro) that LLMs trapped in desktop applications will attribute themselves to patches manually. It may be in a random PR or email patch. Oh, and someone could have given Claude write privileges to a fork.
Look up “git-blame-someone-else”.
You can just turn the intelligence bullshit off on macos, no mee for 3rd party software. I’ve never had it suggest or correct anything on my work laptop.
Not the new one. Maybe don’t update to 27.
To be clear, you had to opt in before. Now it’s just there.
Now, you can turn off Siri. But “Apple Intelligence” is now baked into the OS. All the 27 OSes (macOS, iOS, etc).
Hence the command line tool all the tech blogs are talking about.
The person behind the wheel is more important than whether it was made with the assistance of AI. An actual IT specialist knows what needs additional testing and polishing, so the software will be as functional as average human-written one. (With the exception of IT geniuses like Torvalds or Davis, because most software developers aren’t anywhere near their capabilities to invent.)
You can encounter organic-written software full of bugs, and then sold for actual money. Bonus points if it’s still 32-bit, or even 16-bit and works only on Windows. (Take a stroll on CrossOver forums, you’ll find many still use such archaic software as part of their daily life.)
Let’s be reasonable and not replace an extremism (AI fanatics) with another extremism (anti-AI fanatics). It’s often the case one can find unnecessary hate on both sides of these self-imposed blockades.
As an aside, NovelWriter is a great FLOSS writing app. It’s very much like Scrivener, but it uses Markdown instead of being a rich WYSIWYG editor.
