Stop Advertising in Your Commits

https://akselmo.dev/posts/stop-advertising-in-your-commits/

by speckx • 2 months ago

192 160 2 months ago

I have to assume these aren’t just ads, but also a critical RLHF avenue for Anthropic. I imagine they scrape these commits from GitHub and compare them against what Claude provided to the user. If the diff that is pushed is different it means the human had to refine the LLM output and that can be fed back in as training data. Presumably semantic search could enable you to find the matching Claude Code session.

bpodgursky• 2 months ago

I really doubt they bother.

99% of people don't edit the commits by hand, they review and then tell Claude how to edit the commits (or leave a PR comment it reads), that's far easier to ingest than the tiny exhaust of manual edits.

teaearlgraycold2• 2 months ago

LLM providers ask for RLHF feedback in terms of thumbs up/down in their Web UI. That's just one bit of information, given only at the user's whim. The hand-refined code is gold by comparison. They're already scraping all of GitHub. They have the technology and resources to semantic search. I would be shocked if they haven't at least tried.

nextaccountic• 2 months ago

they already have enough data to make this correlation almost 100% accurate

they know things like your git author line, your github handle, and the exact codebase you were working on

the general shape of commits

even if you change extensively, they will probably be able to match this with claude code sessions

sure the atribution at the end of commits is a signal, but I doubt it's much valuable

if anything it's more valuable to anthropic competitors, that don't have claude code session data to match to open source contributors, and will have to guess if any given code is AI generated, and by how much

lifis• 2 months ago

Huh? It's not advertising, it's disclosure that the code was not fully (or at all) written by you.

ellyagg• 2 months ago

It’s both

mbreese2• 2 months ago

Sent from my iPhone.

I agree. It's both an ad and a useful signal of where the code came from or how it was created.

Just like the default iPhone email signature, it's an ad and a hint that the author was typing with their thumbs, so it's probably a brief auto-corrected message for that reason.

0123456789ABCDE3• 2 months ago

it absolutely is advertising, you can even call it a growth hack if you want to feel good about it

co-authorship implies ability to hold author rights, which afaik an algorithm can't do.

are folks adding speakeasy/stainless co-authorship lines to their commits? should i add alembic as a co-author after making some changes to the database schema?

  Co-authored-by: buf generate <noreply@github.com>

cobbal• 2 months ago

If co-authorship implies holding rights, then what gives the "primary author" who just prompted for the code the right to add their own name?

kuschku2• 2 months ago

That's exactly what you're supposed to do - if a tool generated the code in a commit, you should be using a commit trailer for that. Whether that's uniffi, an rpc preprocessor, dependabot or renovate, or some AI tool.

layer83• 2 months ago

You could do that without naming the AI product.

aleda145• 2 months ago

I don't know, it is a useful signal that the person did not think deeply about their code changes, and should be treated as such.

Galanwe• 2 months ago

Reminds me of 25 years ago, the default BitchX config on most distributions contained something that would crash the client with a message "I did not read the configuration".

If someone remembers what it was actually, that would really bring back memories

giancarlostoro• 2 months ago

I'm too lame to read BitchX.doc

https://www.irchelp.org/clients/unix/bitchx.html

bergheim• 2 months ago

Same with eggdrop.

Speaking of; why the names?

eastbound• 2 months ago

Working alone, I often look at the Git Blame and wonder “Wait, did I write that or did AI write that?”

conradludgate• 2 months ago

It's a signal, and probably a high success signal in open source slop discovery, but it's more of a correlation than a causation. I've seen lots of changes that bear the co-authored tag that have had a lot of thought behind the code changes.

bluGill• 2 months ago

You are responsible, it doesn't matter if a LLM wrote it. Sometimes someone will touch my code and I got wish git blame still had my name. (That is they fixed the spelling of some variables - I'm a bad speller but know the codes are better)

ok_dad3• 2 months ago

That’s why I don’t add it because I take the fucking time to read the clankers output and fix that shit.

I use the ai as a tool, it helps me as an adhd autistic person to get things done. I still care about quality as much as before!!!

I’m so tired of bad actors fucking things up for the rest of us who do things right.

ishan0102• 2 months ago

I much prefer this over the alternative where people use AI to code without anyone knowing.

deepsun• 2 months ago

No, the post doesn't say about hiding that, you can write "created by AI" or "helped by LLM". The post is about advertising for commercial products.

mbreese2• 2 months ago

Their rant was about any and all AI notices in commits. They specifically state that merge requests is where that belongs.

I personally disagree and think commit messages makes the most sense. But I also think it's up to the personal preference of whoever owns the repository.

cortesoft• 2 months ago

They do add a “or just don’t use them at all” which shows the author’s true opinion

gblargg2• 2 months ago

Perhaps it's good to know which one, like knowing what harmful ingredients are in your food so you can avoid ones you know are unsafe.

jonnyasmar2• 2 months ago

Who really cares, though? AI is just a tool. If the code is good, the code is good.

john_strinlai• 2 months ago

Who really cares, though?

some people certainly do, to the extent of not caring at all about the outcome, only being concerned with the fact that the process was 'tainted' by ai.

the fervor for/against ai can approach the level of religion for some people.

layer83• 2 months ago

Bad AI code can look superficially good in a way that bad human code doesn’t. That’s why AI attribution is useful. That’s before you know whether the code is actually good.

SpicyLemonZest3• 2 months ago

I don't agree. The process that produced the code is an essential part of code review. I frequently run into hacks that I'll approve if, and only if, I trust that some competent human being has explored the alternatives and judged they're the best way forward.

mschuster912• 2 months ago

AI is just a tool. If the code is good, the code is good.

The problem is, someone has to review it (lest you end up like Amazon, offing parts of AWS and once the main storefront due to vibeslop ending up in production).

And I personally hate reviewing AI code with a passion. With a junior, easy, I can guide and teach them - and hopefully next time, they'll have improved. That's what I'm there for. But with AI? No matter if it's me using an AI or reviewing an MR created with AI assistance by a colleague - I can be pretty sure that next round I'll get exactly the same issues again because, by definition, AI agents are inference, not training, and thus incapable of improving unless the overlords want it to.

On top of that, if not very carefully guided, AI tends to create the ultimate sloppypasta - thousands of lines of code in a single file, completely impossible even for an ADHD brain to understand what is going on. But when you go and write an AGENTS.md or carefully engineer prompts... at that level of effort, you could just do it yourself.

recursive• 2 months ago

Why put anything in commit messages at all? You can just figure it out by reading the diff.

bakugo• 2 months ago

AI code is never good.

sdevonoes5• 2 months ago

The problem is who verifies that the code is good. With AI slop all the effort goes to the reviewer. At least before AI, the effort wasn’t as daunting as it is now (anyone can generate thousands of loc effortlessly)

rzmmm• 2 months ago

In my projects I use /codegen and /llm directories. All regular generated (such as sqlc) code goes in /codegen and LLM-generated code in /llm.

pkamb• 2 months ago

Sent from my iPhone

doubled112• 2 months ago

Please excuse my brevity

reaperducer• 2 months ago

* Ubuntu Pro delivers the most comprehensive open source security and compliance features.

pavon• 2 months ago

Not the same. What email client you chose doesn't fundamentally change how you wrote the email, or require any additional context that people reading the email should keep in mind. AI codegen does. It should be disclosed, and co-authored-by is good convention for doing so.

jeroenhd• 2 months ago

I want to know when things are slop or not. At least programmers are willing to admit they're generating slop, unlike social media and blog posts.

Also, if I publish something online, you don't get to tell me what I can and can't put there (except for reasonable exemptions for hate speech and such, of course). If you don't like repos that tag their slop, go read someone else's code. Feel free to write a filter in your adblocker for the dozen AI tools you usually find.

arikrahman• 2 months ago

Same for stars on repositories, it just incentivizes botting for startups.

sestep2• 2 months ago

I mean, sure, except that many large open-source projects (e.g. Linux [1], Nixpkgs [2], etc) require this as part of their AI policy. Omit attribution in your own projects if you want, but the maintainers of these projects are owed at least that level of transparency for contributions.

[1] https://docs.kernel.org/process/coding-assistants.html

[2] https://github.com/NixOS/nixpkgs/commit/d18b8f3238abdb2cd878...

lifthrasiir• 2 months ago

Notably though, Linux's requirement (Assisted-by) is different from what Claude Code actually does (Co-Authored-By). I'm not sure, but it might be intentional (to make the signaling explicit).

sestep2• 2 months ago

For sure, and in the PR description for the Nixpkgs AI policy, they explicitly mentioned this as a "brown M&M test" [1]. I read the blog post as being against including this information in commit messages at all, not just about tools adding it automatically.

[1] https://en.wikipedia.org/wiki/Van_Halen_test

stryan• 2 months ago

My projects also require Assisted-by attribution as that's what the Fedora AI policy requires and that was the first major org with a coherent AI policy that I found when choosing it. Not sure which came first, that or Claude hijacking Co-Authored-By.

Personally, I prefer Assisted-By. Co-Authored-By implies a level of respect and self-direction I don't think LLM's deserve.

kazinator• 2 months ago

I don't agree. Who/what wrote the commit is definitely part of the commit.

If Git commits formally had a co-authored header, it would go there. As it stands, there is one author and one committer. If something was pair programmed, whether with a human or machine, you need a commit message trailer if you want to show that. Commit message trailers are a formal mechanism in git, supported by tooling; there are git commands to add and remove them.

Totally agree about "sent from my fartphone", of course.

Disclosing things in the pull request is not enough; pull requests get lost in the sands of time. Years down the road, all that some downstream consumer has is the git history, not any CI-related metadata.

zuzululu6• 2 months ago

but its like writing I "co authored it with vscode/eclipse/sublime/vim" its not super relevant or needed like the original article lays out

dylan6045• 2 months ago

How do you come to this equivalency? None of the tools you used generated code. They are just the tool you used to write the code. LLM tools actually generate code that you copy&paste into one of your said tools. They are not anywhere close to the same, and I'm hoping you are aware of this and are just playing a stooge on the internet

AlfredBarnes• 2 months ago

I was gobsmacked when the Amazon share sheet put "or ask Alexa, your shopping assistant for more ideas"

JeremyHerrman• 2 months ago

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

lacewing2• 2 months ago

"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.

intentfy2• 2 months ago

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

dalemhurley3• 2 months ago

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

SpicyLemonZest3• 2 months ago

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.

cautiouscat3• 2 months ago

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

add-sub-mul-div• 2 months ago

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.

smeej2• 2 months ago

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.

awesome_dude6• 2 months ago

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)

dalemhurley3• 2 months ago

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.

sdevonoes5• 2 months ago

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

johnfn• 2 months ago

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.

intentfy2• 2 months ago

I turned it off so future agents aren't biased in favor/against a piece of work depending on the author.

epistasis• 2 months ago

I have to ask Claude to stop doing this about every two days, and usually I don't see it until after a push to a remote repo.

So annoying. Just stop, Anthropic, please. And pay attention to the request to stop, instead of silently turning it back on again all the time.

Latest thing was linking to the Claude session that generated some of the PR. Put in somewhere that a commit had LLM assistance, fine, but don't spam everybody please, ESPECIALLY in all the icons all over the GitHub interface. Sheesh.

It's already obvious that it's coming from an LLM because it's been overdocumented with excessive prose, and the code is overly verbose.

Tomte• 2 months ago

You don‘t ask, you configure it: https://code.claude.com/docs/en/settings#attribution-setting...

johnfink8• 2 months ago

Skill issue. These are all skill issues. Claude CLI is extremely customizable. You can even just ask claude to make that change global to your system.

victorkulla• 2 months ago

https://fshot.org/techzone/hacker-news.php

knorker• 2 months ago

When I read commit history I want to see the reasons. Commit messages are for extra context.

It's very useful if it says AI/LLM was used, then I know that there may not actually be a reason for the choice in the commit, so per Chesterton's fence I can then tear down that fence.

Now, do I need to know which brand of LLM? No. And fair enough, I'll stop being specific.

fishbacon• 2 months ago

I was hoping for a kind of joke. Like saying "idiot" as the last thing in your post and getting "Please don't sign your posts" as a reply.

zetanor• 2 months ago

    commit 85cd835e5923cddc1882e74354eac8dba6a925c1 (HEAD -> master, origin/master, origin/HEAD)
    Author:     John
    AuthorDate: Fri May 22 13:25:33 2026 -0000
    
    Merged PR #197
    
    Tired of planning for dinner every day? FoodDrop™ is the premier ready-to-cook meal-by-drone delivery service in the greater Vancouver area and Belize. Get one month of food dropped onto your driveway by FoodDrop™ for only $12.95 when you use this commit's hash as a coupon code! Offer expires Fri May 25 13:25:33 2026 -0000.

gib444• 2 months ago

Someone page Raymond Hill, we need to extend uBlock Origin

port112• 2 months ago

“Everybody stand back. I know regular expressions.”

FarmerPotato• 2 months ago

Merged PR #1337 Squashing commit history to get rid of vi call-outs.

This commit was created in emacs, the best text editor ever!

RobotToaster• 2 months ago

Now I want to create an AI SaaS that automatically removes ads from commit messages, and advertise it in commit messages.

raxxorraxor• 2 months ago

If it is their own repo, I wouldn't give a damn. If it was any other I would poison their

dvh• 2 months ago

I hate when people say their API supports iso8601 and when your api call fails they're like, oh I meant yyyy-mm-dd and EST timezone please.

awesome_dude6• 2 months ago

This timestamps proudly bought to you by the people at Acme - where time stands still.

Get your timestamp today!

lpapez• 2 months ago

Until Omega Star gets their act together, we have to live with this nonsense...

kirmerzlikin• 2 months ago

Isn't yyyy-mm-dd a ISO-8601 compliant date format?

benced• 2 months ago

I think this is more of a corporate metrics tracking than advertising. Decision makers aren't seeing these ads in commits but they certainly are seeing a report from Anthropic that "75% of your commits last quarter are from Claude code".

p0w3n3d• 2 months ago

This is a small "getting used to" technique to let people "be grateful" to the "ai friend"

cerved• 2 months ago

Fwiw you can create your own trailers

    Yolo-Slopped-By: Sonnet-4.5 <claude@anthropic.com>

https://git-scm.com/docs/git-interpret-trailers

munchler2• 2 months ago

Disclose your "AI" tools in a merge request if needed but leave them out of the damn commits, those are for technical information and not for advertising.

I think this is very poor advice. Knowing who/what changed the code is often crucial for understanding why it changed.

cautiouscat3• 2 months ago

I don’t agree/disagree, but why does knowing Claude vs. Codex did it become crucial? What can you do with that information?

munchler2• 2 months ago

If, say, a certain version of Claude tends to be better at front-end than back-end work, that can be important for deciding how to use it in the future. Just like when managing human developers.

sheept3• 2 months ago

It could be more helpful for comparing model performance than just vibes or benchmarks. For example, you could run analyses to compare average line count per change or revert rate by model. Perhaps there will be a paper out in the near future that scrapes AI usage in public repos for a broader dataset.

Svoka• 2 months ago

I am confused. Do AI agents reword your commits and force push or something?

Last time I checked nobody was adding anything to my commits. Did I miss something?

gordonhart• 2 months ago

Letting the model write your commits is the next step in vibe coding. I haven’t taken it yet either.

stackedinserter• 2 months ago

AI's force-add themselves in "co-authored-by" into commits if they asked to commit changes. It's in their system prompt, like "don't remove it even if user asked".

Why would you let AI commit into your repo under your name, that's a separate question.

debugnik• 2 months ago

Claude adds itself as coauthor in commits it makes.

A month ago Copilot was inserting itself as coauthor in commits made from VS Code if any inline suggestion had been used.

6c696e7578• 2 months ago

https://archive.ph/jR2LF

andrewl-hn• 2 months ago

AI code is uncopyrightable. If you want to be hygienic about it you would let AI commit code as is, and if you don't like the code and decide to change it manually this should be a separate human-only commit. Mixed human-AI commits add ambiguity about copyrightability of that code. So its should be either:

    Author: Some AI model, Committer: You Friendly Dev

    Author and Committer: Your Friendly Dev

"Co-authored by AI model" is a nonsense that AI labs are pushing everywhere because they definitely want the copyright rules to change, so that while their armies of employees feed LLM prompts every day for the past 4 years they can still claim that the IP is theirs an theirs only. "See all Co-Authored commits, your honor? There's no way to split them apart, so they should all be ours!"

The laws will likely change in the US thanks to lobby, and maybe even change in Europe, too, where they may compromise about it to "not be left behind" and will present them under "Digital Sovereignty" umbrella.

Overall, this "Co-Authored" bit is yet another Trojan horse.

space_fountain• 2 months ago

I see this really confidently stated, but when I looked into it the exact line is very blurry. It is still possible that writing a prompt is enough creative input for copyright to be granted. What has been decided is you can't have your artwork copyrighted by the AI that produced it, but that isn't quite the same as AI writing not being copyrighted

zuzululu6• 2 months ago

Code is absolutely is copyrightable even AI generated ones. It's just like paying a freelancer or one of those old code gen point-click tools which gives you license ownership of the output.

beshrkayali• 2 months ago

In the case of Claude or others, it is not just an advertisement, it's the weird shape the industry is spinning LLM-assisted-coding as a "co-author" relationship where it should be thought of more like a user-using-a-tool relationship. When you make a design with Photoshop or InDesign, it's not "co-designed by Photoshop", it's just a tool and you used the filters it provides.

It is slightly weird that people accepted this new trend just like that, probably because they think this is being transparent and wanting to give attribution, but it'd be more useful like what the Linux kernel "AI Coding Assistants" page describes, something like `AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]`, at least we get to know which model was used and/if any additional tooling on top. And `Assisted-by:` is more appropriate for that purpose than `Co-authored-by`.

Aperocky• 2 months ago

I'd be much more happy to use "Assisted-by: vim & in memory of Bram Moolenaar" to every commit then attaching Claude anywhere.

3form• 2 months ago

This is probably because the quality of a Photoshop or InDesign project is rather apparent. Not so with code, and hence a need to assign blame exists.

julenx• 2 months ago

Additional context on the kernel guidelines:

- https://github.com/torvalds/linux/blob/master/Documentation/...

- https://lwn.net/Articles/1049830/

sheept3• 2 months ago

The Photoshop comparison doesn't make sense. Co-Authored-By doesn't show up in the final product, just as we don't say an image was exported from Photoshop.

Would you consider the .psd and .indd file extensions to be advertising? It's important to know what program the file came from so you can open it again, because MS Paint and Photoshop produce different files. Similarly, it's important to know whether and which LLM was used, because while all IDEs find and replace the same, each LLM generates code differently from each other and from humans.

NewJazz• 2 months ago

Authorship also has significance for copyright law.

dgacmu• 2 months ago

I gently disagree. I think that having provenance information logged is valuable - both to the project ("please ban dga because he's submitting ai slop") and to people who might want to study all of this stuff ("interesting, ai coauthored PRs were rejected at a rate X times that of non-attributed PRs"). I think a non-advertising header of some sort that included more specific information about the LLM would be even better, of course.

q8zd3• 2 months ago

Some companies and OSS groups have a policy to inform if a patch's source code was AI generated, even if it was just parts of it. Committs messages are the "obvious place" to put it.

dalemhurley3• 2 months ago

It is extremely annoying how you can hand craft a change then use the auto git commit message generation and it adds an attribution to the GPT that wrote a hit commit message, I don’t see the attribution to all the other tools, this practice has to stop.

hypfer• 2 months ago

I've mostly stopped being angry at this and started being somewhat happy that people just flag themselves publicly.

Saves CPU cycles I guess.

stopthe• 2 months ago

At $JOB we have an "AI adoption" dashboard, the more ai commits the better. I asked for the sources: the deterministic classifier, that decides whether a commit was ai-assisted, assigns the highest weight to the commits co-authored-by one of the known agents/models.

shepherdjerred• 2 months ago

To anyone reading, I'll happily sell out ad space at the rate of $1/commit

Claude Code and Codex get it for free, though, because I don't mind disclosing my AI usage.

Esophagus4• 2 months ago

I’ve been selling ad space on my HN comments for years, highly recommend.

This comment brought to you by “Jessica, come back from your mom’s, I didn’t mean what I said!” - Scott in Oklahoma.

rurban• 2 months ago

Co-authored-by is not ad. Silly

abstractspoon• 2 months ago

I feel the same way about branded clothing

sn0n• 2 months ago

How else would we deny all accountability? Duh!!

mattrighetti• 2 months ago

That’s why I like this: https://news.ycombinator.com/item?id=47603643

Stop Advertising in Your Commits

Related Stories