[JeremyHerrman]

Is AI responsible for the committed code? Should AI be blamed when services go down due to the change?

The answer is absolutely not - the developer is responsible whether the code was AI assisted or not, and the dev's name should be attached to it just like any change.

The OP is right: these are ads, plain and simple, and it's a dark pattern for these companies to have attribution enabled by default

[johnfn]

If a particular AI provider is responsible for a disproportionate amount of buggy code I absolutely want to know that. Perhaps we'll switch providers. Sure, this will be flagged in CR too, but I would want to see trends over time.

I don't know why you wouldn't want to track this information.

[SpicyLemonZest]

I don't agree at all. I think it was a dark pattern when attribution was not enabled by default and angrily demanded it from everyone who sent me AI-generated code. It's not OK to declare sole authorship of something that you did not in fact author simply because you're volunteering to take responsibility for it.

[cautiouscat]

The dark pattern was adding the name of the product. An acceptable pattern would be “Written with the help of an LLM”.

[neilv]

I disagree. Imagine a human co-author on a work. "Written with a co-author" would not be sufficient attribution.

[cautiouscat]

That’s Apples to Oranges. Me saying “Co-authored by Joe Smith” gives the human, Joe Smith, possible exposure and definite credit.

“Co-authored by Copilot” gives a multi billion dollar corp free advertising. I don’t care about them. I do care about Joe though.

[dylan604]

Knowing it was Copilot vs Claude vs ChatGPT makes a difference just like knowing it was Joe Smith vs John Doe

[groby_b]

It really doesn't. Same as we don't say "written with vi" or "written by Emacs", even if it is intuitively clear one of the two is better.

[bigyabai]

Just like how you know that Brawndo™ is the different, better version of the pedestrian "water" that everyone drinks.

How could it possibly be product placement? It's got electrolytes!

[neilv]

Positive credit is not the only purpose of attribution.

There's also the named authors not taking credit for something they didn't fully do, regardless of whether the credit goes to someone else.

There's also traceability, if the authors/provenance needs to be considered because of some kind of problem or potential problem -- technical, legal, security, or otherwise.

[bigyabai]

> There's also the named authors not taking credit for something they didn't fully do

Advertising in git commits is not ever going to substantially discredit these people or hold them accountable.

[smeej]

Your caring about the entity cited doesn't actually change the nature of the citation. Your saying "Co-authored by Copilot" does the same thing--gives Copilot possible exposure and definite credit--even if it doesn't need it and you don't care about it.

[ok_dad]

A clanker isn’t a human it’s a tool. I don’t write “Coauthored by VSCode” when I use find and replace.

[cperciva]

I write "reported by gcc16 -pedantic" when I'm fixing a bug, because it's useful to flag "this is how they issue became visible" in case readers want to use the same tool on their code.

The fact that it's just a tool is irrelevant. You don't need to mention search and replace because everyone already knows that exists... mind you I have had commit messages which included sed commandlines, for ease of future reuse.

[awesome_dude]

I know this is a bit off tangent - but can you please convey that point to every damned dev/team that ends up with a freaking ".vscode" directory in their repo

[1718627440]

I use the GNU Autotools toolchain to generate VSCode folders. We are not the same!

(The first sentence isn't a joke.)

[bulbar]

I have done that in the past to provide zero-setup development environments for our internal python packages.

What's the problem with the approach?

[sodapopcan]

They don't know what a global git ignore is and if you tell them they don't care.

[sheept]

Find and replace behaves the same in most IDEs, but for large changes, the specific LLM can affect the generated code for the same prompt

[ok_dad]

So does my brain, but no one complains. If I’ve read and understand all of the code it’s effectively the same thing. I’m washing the LLM output so it’s clean. My brain parses the LLMs code and deems it safe.

[tadfisher]

Why does the LLM need the same attribution as a human? Feels like a false equivalence.

[SpicyLemonZest]

I get the opposition to product names, but as someone who's trying hard to make an OpenAI boycott work, I personally value being told which LLM in particular is responsible.

[awesome_dude]

When a paper is submitted to a reputable publication references are demanded.

You have to let people know where your ideas are supported, or even come from.

To do anything else is plagiarism.

AI isn't a co-contributer - but it should be referenced - just like a link to a Stack Overflow comment when that's the source of code.

Having AI referenced in the commit is (IMO) best practice - but only co-contributer attributes are available (for now)

[dalemhurley]

Disagree, I never attributed Laravel Artisan or all the other code generators.

Right now you could run Laravel new app (replace with any new framework) and Claude/Cursor/Codex git commit will claim to of created the code.

[awesome_dude]

To be fair I've never used Laravel - but I do have a memory of code generators/scaffolders inserting files that made clear they were responsible for work

/*

* NOTE: This class is auto generated by OpenAPI Generator (https://openapi-generator.tech).

* Do not edit the class manually.

*/

[sdevonoes]

Do I need to write “coauthored by protoc” when pushing grpc changes? “Coauthored by gofmt” when formatting? “Coauthored by React when using npx create-next-app”?

No. So, same goes for ai

[awesome_dude]

Have a look at the top of the files generated by protoc bud ;)

It's not in the git commit, it's in the files themselves - is that what you want for AI instead?

[add-sub-mul-div]

It's not the gun's responsibility when you shoot it but we still need to have discussion and rules about guns. Enough with this tired, worthless semantics argument.

[smeej]

And there is something useful about being able to trace the ballistics back to find out which gun was responsible for the shot, as a key to who was wielding it and is ultimately responsible for its use.

[lacewing]

"Sent from my iPhone"

But of course, owning an iPhone early on was seen as prestigious. Using an LLM is... not? Many people really don't want the world to know. For blogs in particular, the urge to have an LLM generate the entire thing and then post it under your name seems to be really difficult to resist.

[intentfy]

If you're turning commit attribution off, you're definitely turning iphone signatures off too.

[dalemhurley]

Nice deflection there, why can’t both things be bad?

[lacewing]

I don't know what you're saying.

This is the second or third HN post I'm seeing this month along the lines of "how dare AI companies flag my code as AI-generated". I just don't remember similar complaints about the iPhone footer. Not many HNers complain about The North Face putting the text "The North Face" on their hoodies either. Or Honda putting their logo on the car.

The reasons for this difference are interesting. The fact that companies put their logos / brands on stuff is a lot less interesting to me. You can call it bad, but again, why is this instance worse?

[port11]

Plenty of people turned off the iPhone thingie once they figured out where the setting was (kinda buried).

Plenty of people don’t like logos. And some do, because they tie their identity with brands. Maybe in the future some of us will tie our identities to specific agents/LLMs.

To each their own. I don’t have clothing with logos, I find it gross. And you also don’t know I’ve sent you an email from an iPhone. Ça va.