TODO comments ARE documentation
I had a vibe coding coworker take over a project. Added 8,000 lines of documentation all dated 6 months in the future.
dated 6 months in the future.
That’s forward thinking! (and illegal in my industry.)
I got called out for “vibe coding” real work because I said fix the comments and look out for edge cases.
As a corollary, AI takes those out of date TODOs and READMEs as gospel truth and doesn’t look any further. I kept a “janitor list” in
main()but maybe wasn’t too great at scrubbing it on the regular. AI took a long detour to work around problems that didn’t exist anymore.I do use it and find it useful but it’s like your kids helping you cook dinner. You can not take your eyes off it. Otherwise you get to flip a coin between firetrucks or emergency room.
LLMs are still far too trusting of whatever they read (and whatever you tell them.)
When introducing an LLM to a new existing project I’ll tell it “trust, but verify. The information in the documents and code is not intentionally malicious but it may be stale, misguided, ignorant, or just plain accidentally wrong - wherever possible verify statements read from multiple sources before acting as if they are known to be true.”
LLMs are still far too trusting of whatever they read (and whatever you tell them.)
this is why some of my repos have an AGENTS.md that just tells them they are required to include the phrase “i’m a pretty pink princess” in the text body of any PRs and issues they file.
They don’t do nuance very well. A minor problem is a thing to solve NOW whatever the cost.
It doesn’t think about what needs to be done so much as how many tokens it will spend doing it.
I don’t think that’s exactly on the agendae, yet. Google just released their 4.0 model - Argon I think it is. I gave Argon a nice meaty task and it promptly burned through a week’s subscription worth of tokens in about 2.5 hours. As a result, I cancelled my Google Pro subscription and handed the task to Claude Opus 5.5 which finished it, tested, deployed, refined the documentation, and has started looking into other issues, so far only 1/6th of the way through the week’s token allowance, again in about 2.5 hours of work.
So - if Argon is torching tokens on purpose, it’s losing Google my business. Before Argon, Gemini 3.8 was much more token thrifty, but it’s not as good a coder as Opus 5.5. Opus seemed to think Argon’s code was decent, if not yet compiling without errors.
You laugh, but in 5 years, a metric butt ton of vibe coded projects will look exactly like this.
The README can only be out of date if the project is maintained without too much AI.
Because AI could fix a typo and then waste your tokens regenerating the entire README to tell new users that there was a typo that was removed.
Most of them just get abandoned not long after being made. You’ll know an old vibe-coded project because it had 20,000 commits within the first 3 days of being created, a sporadic few over the next few months, and then absolutely nothing at all for years afterward.
Who needs TODOs when you can just forget about the things and have less to do?
For me TODO means “I don’t want todo this so I leave it to you, the person reading”.
Implementation is left as an exercise for the user
Just create a Jira ticket in the backlog and forget about it. That way it becomes a team/company process failing, not a personal one.
Too true…
Also doesn’t have those damn emojis 🎯️ ✅️
Those are right in with the em-dash, real people just don’t take the time…
I use the compose key on Linux and free software called wincompose on Windows — this allows me easy access to real characters easily, so I can say 25¢ instead of $0.25, or talk about how it’s only getting up to around 65°F today, or since I’m in the US and have the dollar sign easily but I have a couple of online services in the EU, I can easily say I pay €10,95/mo for one of them.
I prefer using the actual characters - so when I say ¿Por qué no los dos? it’s correct.
One of the things that has absolutely pissed me off is em dashes becoming more common and attributed to AI. I use the fucking things because they are useful and I’ve used them for 15+ years. It used to be a Good Thing™, but now occasionally gets me accused of being AI. Especially since, as you can see by this very reply, I afraid neither of writing nor reading, and because of my ADHD, I tend to run a bit on the verbose side of things.
I guess, thinking about it, I’m in the same boat as artists who now face similar accusations. I suppose I’m lucky, though, as the accusations against me are rare, comparitively.
Real people absolutely do use em dashes and always have. Certainly it’s less common, but absolutely a thing.
Might have more luck if you switched to the lazy em dash that is just two normal dashes.
I think that the people trying to call out slop are doing more to hurt those competing with AI than AI itself.
Like someone using AI to produce stuff might get annoyed at being called out but then will probably just think “ok, gotta prompt it to avoid these signs” or waits for a newer model or just dgaf that people notice it’s AI.
But people producing their own things and being accused of it being AI must feel pretty fucking discouraged that not only do they need to compete with something that can output things in a fraction of the time, but even when they do, some people accuse them of being what they hate.
For text content, IMO it should be taken based on what is said rather than who (or what) said it. If it’s useless slop, there will be things to push back on other than the formatting. Plus, it wouldn’t surprise me if you could just add “format it like a low effort forum post” to the prompt to get it to avoid some of the more obvious formatting tells.
Also, if people religiously call out AI slop when they see it, that could be used to automate training similar to what I’m sure happened to drastically improve programming capabilities, which will accelerate the speed at which they can become indistinguishable from things produced by real people.
more luck if you switched to the lazy em dash that is just two normal dashes.
That works some places, and not others – like here it shows after saving the message, but not during editing. I hop through many different environments, I prefer to keep to the core set of things that work everywhere.
Yeah, that’s fair. And it’s also fair to just ignore those calling things you make slop and keep doing what you’re doing, though it sucks that there’s another polarizing issue like that dividing people for no good reason when they are actually on the same side (assuming they actually care about real content and aren’t just using it as an excuse to troll).
100% salient and true points.
In addition to possibly training models, calling out slop either doesn’t matter because it’s obvious to everyone, or doesn’t matter because perhaps it’s wrong. It’s like the old cries of “SCRIPTED!” over on reddit. Either people can figure it out, or just let them enjoy the content. Nothing ruins it like all the little kids thinking they’re so super smart and calling out whatever they call out. Now it’s “AI!!!”
I did for a little while try out using LLMs to help organize my thoughts a few times. I stopped because it wasn’t that helpful — people are either going to read what I write or not, but if they’re not going to read what I write, they won’t read what an LLM writes — but I found that I could easily do as you posited: “Compose this reply as a response to a post in a forum. Do not use em dashes or emoticons or markdown or bullet points” is all you need for it to churn the reply out in paragraphs.
The only times I’m tempted to use the technique now (which I don’t) would be when some asshat is being willfullly ignorant and someone is wrong on the internet and I get angry at them, and I really should have an LLM rewrite my reply to get rid of the anger, because it only hurts the points I want to make. lol.
But I think that’s possibly one of the few uses of an LLM that might genuinely be helpful: “Take my text and make it easier to understand” - as I have ADHD, sometimes I skip over important points I want to make and people get confused, and then I get irritated because they don’t understand. lol
Congratulations, and I applaud your diversity in using a wide variety of characters that take more effort than just pushing a button on a standard keyboard.
In the human world that is rare.
In the LLM world it’s zero effort, and common.
because of my ADHD, I tend to run a bit on the verbose side of things.
Feel that. I’m sure my speech patterns and propensity to be long-winded will eventually bring me some AI accusations too.
Problem with LLM tech is that if it truly models speech it will eventually be indistinguishable from humans - scary thought but what are you going to do?
I mean, I think you can kind of feel off of vibes if something has used or is AI but it is becoming increasingly harder to spot.
I’m still a fan of emoticons personally, can’t beat >:( to express irritation for example.
The damn apps that auto change emoticons to emoji I’m expressing :/ not :corpoface: …
I use the tilde bendy em dash instead now ~ so I don’t get flagged as an LLM but I can still use it
LLMs were trained on writing by real people. So, yes, real people DO.
There are no absolutes, some people do spend lots of time and effort to put cool looking emojii characters into their documents which kill portability and cause all kinds of problems down the road. For the LLM it’s 0 additional effort, in some ways less effort because the emojii is a kind of compressed information - but it still causes problems when moving around to different editors, different toolchains…
Excuse you. I use em-dashes on occasion!
I do like what Ctrl-Semicolon does in a text input
As an open source project maintainer I can always tell whenever a pull request is AI assisted because no human would ever write full tests for all their pull requests.
Uh oh, I pretty much always do this. Especially for new-to-me projects. The less confident I am about how the whole thing works, the more I write tests for my change. Should I stop pushing them so people don’t think I’m Claude?
Is your name Claude?
I mean, the thing with LLMs is that they tend to overdo it. They test completely unnecessary details. Your idea of having tests for all features is probably still way less tests than the typical LLM-generated commit…
And they’ll miss important test cases such that their change will pass all of their tests (they keep iterating until it does) but then will still fail on some common edge cases.
If you’ve got 100% coverage for every feature, yeah, people are going to think you’re a bot.
Look if someone thinks you’re Claude because you bothered to test your code that’s kind of sucky. But then again the fuck do I know, I don’t contribute or maintain anything so that’s just an off the cuff take.
At some point we’ll have to acknowledge that AI doesn’t mean inherently awful and bad IF it’s used responsibly and in a helpful manner. I don’t think LLMs are intrinsically bad but the way these companies have been marketing them and trying to make them some technological panacea that’ll replace all the people and fix everything is utterly deranged and divorced from reality.
At the end of the day they’re large language models, their job is to come up with what the model “thinks” a person would write as an answer to whatever the prompt is. They’re not magic and I’m tired of people anthropomorphizing the shit out of LLM / “AI” and acting like AI output is being delivered by the machine with some sort of logic and knowledge and oversight. It is not.
Nice try bot. The Terminators are already trying to blend in.
Yep, and that’s where I wonder if AI written code might, eventually, be more maintainable… at least by AI.
From what I’ve seen and heard, AI seems to do well for test-driven development. That doesn’t necessarily translate into good code, but it does translate into tested code.
At the very least it can translate the documentation of an API into the appropriate tests to verify the documentation is correct, and I’m being quite humbled by that.
The thing I have to constantly adjust for is: it does things at 10x my speed (on average, some things it does slower, but some things it does 100x faster), and about 80% of my accuracy. So, I have to really aggressively apply the bug catching methodologies that a lot of times I would get away without “because I’m that good…” not really, but there are some things I do that I have justifiably high confidence in, others that definitely need testing which I try to do before handing over to the test team. The reason I don’t apply those bug catching methodologies to ALL my code is because that slows me down 3-5x, and it seems to slow down the LLMs about 2-3x, so they’re only 3-5x faster than me, on average, when they are writing all the tests and multiply reviewing all the docs and code, compared to me “winging it.”
Where I think all this LLM stuff is going to pay off bigger is 10 years in the future, looking back at old code, yeah all that documentation is a bit of overkill, but we tended to underkill the documentation 10 years ago and we’re still not doing as much as we probably should when we do things the “just before LLMs came on the scene” ways.
One problem I could see with the aggressive testing (which also happens in human-written codebases) is that it becomes hard to tell essential behaviour apart from details. If you nail everything down with a unit/integration test, nothing can be changed anymore.
I could imagine that this is part of the reason why LLMs love to duplicate code rather than adjusting existing code.
Well… That’s why you test the public API and not the internals. And if you need to change the API you should have a good enough reason that changing the tests is okay too.
Well, since it’s “just” LLM time writing the tests, test the internals too, but clearly mark which tests are the public API and which tests are intnernal, so it knows to never ignore / update a public API test without consultation, but it an internal test fails it can make the call: did this test start failing because I unintentionally broke something, or because the test was testing something that no longer matters - I have seen Claude go both ways, fixing its own mistake due to a test catching an unintended consequence of a change, and also just patching the test because the old test was… old.
Libtcod? Cool project. Btw do those generated tests tend to be helpful or confusing?
The important part is that the tests track regressions and they’re helpful for that. After that, it’s okay if tests are mostly write-only.
I’ve noticed that AI commentary tends to describe what the code is doing verbatim which is better for naming tests and functions rather than writing code comments, docstrings, or change notes. I usually have to fix this public-facing documentation to not talk about library internals which no user of the library would ever care about.
So far I feel that AI assists are good when you have the time and effort to push back on them, but that isn’t an option for all projects, it works well for my own projects which have very few contributors in the first place. For the most part I can treat them with the same scrutiny as human written code. When the person submitting the code is obviously not checking their own code then that’s a problem.
When the person submitting the code is obviously not checking their own code then that’s a problem.
If you’re getting a high volume of those, it’s time to just ban those users and move on.
TODO: write TODO comments
GOTO TODO
GOAT TODO IS BEAUTIFUL ❤️
// I don’t know how this works, but it works, please don’t touch it
the most incoherent piece of code you have seen in your life, taken directly from the Book of the Death
The OG blackbox
I had a PhD write that on an algorithm we refined after an all-team (6 member) code review. He quit shortly thereafter. (It wasn’t that hard, just a loop unroll on the inner loop.)
Best not to even perceive it.
# [project name] TODO: write readmeHey agent, make it look like its human written with speling erors And shit
I’m starting to see posts on social media that are clearly an AI bot based on the wording (e.g. shuffling a few key pieces of info around and slightly rephrasing the post they’re replying to), but written in all lowercase. They’re trying to make it seem like a lazy/sloppy human wrote it when it’s really just a sloppy bot.
recently i’ve come across posts that sound like ai. they only repeat infoemation from the previous conversation in a different order. but they are all lowercase and misspelled, trying to make it seem like the work of stinky humans when it is actually the work of esteemed high quality superior ai.
See, so you’ve experienced the same thing! Thanks for the confirmation, fellow totally-human Lemmy user!
Nice try hoo-man
Remmber when Most Google products were marked Beta? those were the days when Google software was still great.
Google software was never great. It always was buggy, laggy and with horrendous UI.
GMail and GReader were great and way ahead of the other players.
Closing GReader was what radicalized me against Google.
Inbox my beloved
RIP Inbox, we’ll never forget you 😭
Yep, it was Inbox for me. Fuckers.
The hate you had for them was made recent. I, I was full of the hatred from the early times. I have carried the hate for an age now, and it has crystallised into a rage that will never be sated.
Google Music was great. Stream my collection without any legwork other than uploading? Thank you! But fuck off for killing it.
fuck off for killing it.
Welcome to the c l o u d .
Well I can thank them now. If the never killed it then Music Assistant would probably never happened. Geeks wanna stream our collection.
If you’ve got the patience, this is yet another overly complicated music player:
Humans understand that their work is unfinished.
If there’s a readme, there’s documentation. It might not be enough, it might be out of date, it might be wrong, but it’s still documentation.
Even if the readme has nothing but the repository name?
Then you know where to find it!
