I may be missing something, but this is exactly what I want as a default. I want the session links there. I see PRs from my coworkers with session links.
I still get to control whether other people can see the session, but I don’t lose it.
I don’t get the “unprofessional” argument. This is attribution. Attribution is professional. If you don’t want it, you turn it off or rewrite the commit messages. If you are having Claude write the commit messages FOR YOU and you are NOT READING THEM then that’s what’s unprofessional. I am honestly kind of shocked and disappointed by how upset people are by this.
Which is actually why I DON'T like the session amended. Ultimately it is an ad placed into the Git log, and noise at that. Regardless of if it exists or not, the professional responsibility for what is contained in that PR/commit falls to the feet of the developer contributing it.
Microsoft previously did this with Copilot, which was correctly met with a very negative response. I don't see why an ad by Anthropic is better than an ad from Microsoft.
Agree, that it is an ad. Also, unless the full session (in a resume-able format, along all session artifact such as intermediate research/docs used to produce the final output) is also included in the commit, the session id is useless.
I presume this is Anthropic trying to their "usage" stats before their IPO.
> Anyone who did and could do so turned it off immediately
I know I’m not alone as someone who’s bothered by the postscript, but too lazy to do anything but delete it every time. There’s even an entire (well part of) song about it: https://www.youtube.com/watch?v=EbdeVhPAbms
Out of curiosity, do these Claude commit messages fall into the same category of "it's annoying but not worth the effort to turn it off"? If yes/no, why?
Most people's identities and self worth are tied up in the material things they buy rather than actual self actualization and personal axxomplishments.
Even as a lifelong Apple user and longtime iPhone user, that sig irritated the shit out of me and was instantly disabled whenever I set up a new device.
Also, and this may come as a shock to you and a number of other HN users: I never considered it a "point of pride." I switched because Android was a privacy dumpster fire comparatively and battery life was far better.
I also can't remember a discussion in person with folks where someone seemed "proud" to use an iPhone. If anything, the Android crowd loves to sneer at iPhone users and look down on them as idiots who need a "simpleton" phone.
I'm no simpleton. I ran custom roms and so on. Had F-droid installed, and so on. I do miss the customization, but I also look back and realize I wasted a massive amount of time on installing roms, backups, and the customizations themselves. Probably weeks of my life if not more. My iPhone? I don't. It's a phone. That runs apps. Do I wish I could install whatever I want, particularly given how secure apple's sandboxing is, and thus in theory how safe it should be? sure. but at the end of the day I want to live my life.
Now compare that to why people don’t want Claude in their commit messages or PR descriptions.
Clearly there’s no pride in using agentic coding (if anything, maybe it’s the opposite of pride..). I don’t think most developers want to publicize their use of agentic. But I could be wrong.
I'm not sure i see it as an ad. "Sent from my iPhone" is an ad because the iPhone is not a meaningful part of sending the email. Claude is a meaningful part of the commit though if you're using it, so to me it's more like clarity or transparency than an advertisement.
You don't see how a link to "https://claude.ai/.." in the Git log is an ad for Cluade/Anthropic? It feels exactly like "Sent from an iPhone" to me, and has exactly the same value to people reviewing the Git log later (i.e. none).
Keep in mind that if you run across someone else's spam in the Git logs:
- You need a Claude account.
- You need permission to view the session.
The only people who can just follow that link and look at how Claude was used, is Anthropic themselves. Therefore, they can also use it as an additional mechanism to tie sessions back to other IDs/identities. For everyone else this is "Sent from an iPhone" but longer.
But presumably everyone in your company/team is using Jira, so it's not an "ad" because it's a product already used internally. Claude is appending these links to all commits by default, whether or not others on the team use Claude. Those are very different things.
Good counterpoint. We link to GitHub from JIRA, and the reverse. Linking to the Claude session that generated the code seems reasonable...it's a related artifact.
One thing I'm not sure about: what is the utility of linking to the session? What are folks expected to do with that data?
You're missing the pattern of behavior from Anthropic. They constantly market like this: the session link, the Co-Authored by Claude, the unwillingness to adopt AGENTS.md. Hell, on workflows on 2.1.246 Claude's been ignoring the "attribution" block in settings.json and gone right back to coauthoring every commit.
What's the odds on them fixing that bug anytime soon.
The part after "https://claude.ai/" is a UID, that's useful to have to tie together multiple possibly unrelated commits and changes authored in a given session.
The way I see it, the conversation transcript is the important part. The specific link to claude.ai nor the model itself is important to me. Otherwise we'll all just be linking out to 10 different AI model providers websites to see history. I'd rather it be some open solution.
Models change so much that I don't really see why anybody cares what model was used to generate the code after the fact. It changes every few months and they're non-deterministic so its not like you'll ever "recreate" the same conversation yourself anyways.
UID itself is what is useful, regardless of whether or not you have access to the actual session transcript.
One could argue that "https://claude.ai/<uid>" is slightly more verbose over "CLAUDE_UID=<uid>", but the former is also slightly more convenient for those with access to the transcript, and both have word "claude" in them, because you need to be able to tell which UID in the commit message is for what.
There has been more than I project where I literally didn’t even look at the code.
I prompted what amounts to verbose “I want X” and then did QA. I’m perfectly happy with an LLM being credited with that code and there being no copyright existing for that code.
For some people that is their life’s work and they want to clutch pearls. Their sole contribution was a half baked idea and they shipped QA out to their “users” but they want to be credited with any rights.
There is definitely a continuum in LLM usage for code, do I agree with an AI blindly stamping every commit, not really.
Do I see LLM generated code as equivalent to me digging a hole with a shovel, it depends. Sometimes there was very little shovelling, other times there was no me involved and you may as well credit the entire hole to the shovel.
Where this will come in is when it comes to copyright and licensing.
If your entire product is LLM generated you sure as hell have no right to place any license on there other than potentially a MIT equivalent license, even that is dubious though but from an organisational standpoint MIT probably will be more widely accepted than saying it is uncopyrightable.
My argument being I would rather know that you didn’t even look at the code, which is what this will achieve, whether I agree with it or not at the very least it will make it easier to sift through the garbage.
This is a problem with copyright in general. There's no longer any requirement to provide information about copyright status, so there's no way for the public to know whether something is copyrighted (no doubt the uncertainty is intentional as an additional way to undermine the public domain). This is just as true for expired copyrights, or material placed in the public domain intentionally.
The solution isn't to add notices to non-copyrighted things; it's to require notice for copyrighted things, including date.
Individually, of course, one should simply assume things are not copyrighted if you have no way to know, and do whatever you want. If the author can't be bothered to tell you, you have no moral responsibility to care.
Yes it’s the same as using a stock drum pattern in a drum machine, it is not copyrightable because no human was involved in any part of it.
If you have an idea and want rights to it then do it the old fashion way and pay a developer to do a commissioned work and you can have your copyright, heres the catch, that will cost a hell of a lot more than 100usd a month
And when that developer copies and pastes useful code snippets from the internet into this product you’ve commissioned them to create? Also not copyrightable?
I agree. But I think I would also like the OS to put a fingerprint (I.e committed on macOSX in /usr/marc/… while he was watching porn on xxx). This way we can get the full context by default and better understand the PRs.
I'm sorry if this is a naive question, but why can't you just edit the commit message however you like?
Or, for preference, write the whole commit message yourself? It confuses me a great deal to see projects where someone put in a lot of effort to write a useful and detailed prompt but absolutely couldn't be bothered to write a few sentences for the commit.
Realistically, what do you actually get from viewing people's sessions? I honestly don't really understand why people care.
The specific model used can't be important because theres a new trending one every couple weeks nowadays. The conversation could be wildly out there and have 10+ different iterations of the same thing in it that led to the final product, etc.
It's not like before AI we were asking people to explain every single iteration they went through in their commit body, or to see their entire internal dialogue.
I don't really understand people's hype around needing to see the conversation they had with their AI. As long as the code itself is self-documented and final decisions are relatively well-documented future agents and humans will pick up on it much faster than reading the entire transcripts anyways.
That's not even bringing into the fact that there are people who use multiple models, multiple sessions, multiple harnesses across the same commits/changesets. How do you reconcile those alongside people who could have made human changes in the end?
There’s always been a school of thought that is in this direction. In the pre AI world people discussed whether to squash commits or to not and I was on the squash side because I think that the development artifact should be optimized for reading and just as one does not read a novel in all of its pre-edited forms to read it, one should not do so for code. The cost to grok using that method is too high.
However with LLMs you can have them analyze prior transcripts to identify bug sources and so on. I could see it being used. But all theorycrafting. And perhaps motivated reasoning. I’ve always committed early and often and used bisect to find bugs. That places me right away in the squash commits camp because of the method I use to write code.
The link to session is a red herring - most people seeing it won't be able to access the transcript.
It's a tag with an UID. It's useful to connect together multiple changes, including possibly several commits across multiple unrelated repositories, as well as other artifacts and actions. It records the causal link, and connects it to telemetry (people do run their own telemetry over agents, especially in team/company setting). Because chances are that, if you have issues with some change made by AI, being able to look up what other changes were made as part of the same session is going to be very helpful.
> That's not even bringing into the fact that there are people who use multiple models, multiple sessions, multiple harnesses across the same commits/changesets. How do you reconcile those alongside people who could have made human changes in the end?
This is exactly how. By tagging changes with session UID, and using it as key to resolve other changes, as well as specific model and harness versions used, if you use telemetry endpoints harnesses tend to offer.
The whole discussion is just people getting tripped over the tag being link-shaped, ie. "https://what.tld/<uid>", instead of "WHAT_ID=<uid>".
People don’t want their incompetence displayed. The session is probably a series of “continue” prompts with basic understanding demonstrated in the initial ask.
I think it’s foreseeable that if you use the LLM to commit work you’ve written, it might cause problems with attribution.
In general I assume that attribution to AI is approximate. People copy-paste things out of chats and those don’t get attributed to the AI. And on the flip side, I’ve had AI write a commit, and then I’ve reverted it and written something different by hand (and it got tagged with an LLM session).
The AI attribution is approximate and informative. The commit message and authorship are the more important parts.
You may be concerned that people will think your commits are LLM-generated. I’m sorry that you have to work in that environment, but I don’t work in an environment like that and I don’t share your concerns.
How would it know for certain it was used for a given commit? What if there was no code? What if the approach was different, but some lines were the same?
I think some folks think that using an LLM is a shameful act and this is somehow shaming it and others believe that it ties back to the conversation. These conversations don't have any guarantee of being accessible over the long haul. If there was value in the conversation, the conversation should be somehow captured as well, not just some URLs.
I'm OK with the attribution, but I just didn't like a session URL appearing on a public repo all of a sudden. I was left wondering "did I just leak my private session?"
I don't understand why it isn't opt-in. Or at least a heads-up somewhere.
At this point it's believable to me that Anthropic might not even be aware of what features get added in a given release. It's hard to tell the difference between "they actively are against documenting all of the defaults they keep changing" and "they genuinely don't even pay enough attention to notice when their vibe-coded changes have changed a default". Functionally they're the same, and both would stem from similar (lack of) values, but I think it does kind of matter because it's essentially the difference between explicitly crafting an experience for users versus defining same things they want and letting the vibes end up driving it towards a bunch of user-facing emergent properties that no one has considered.
"Of course they were intentional in adding in an ad to everyone's git log" honestly night be the less cynical take, because the alternative is assuming that they actually look at git logs ever rather than only having Claude deal with it.
I think a lot of people interpret it as trying to foist additional burden on reviewers. As in "here's my PR, I would like to take credit for this idea, it's now on you to review. If you find a problem with it, I expect you to call out exactly what my incorrect assumptions were, which you are able to do because they're all buried somewhere in this Claude session. Let me know what they are and I'll paste them into a new session and submit again."
I agree that attribution is professional. What I would worry about is people using the link to supply the reasoning behind the change without putting it into the commit message. Then you have the problem of needing to load the session to understand the change, when one would ideally be able to get that context from source control alone.
More like an ad for trillion dollar corporations. Not in the business of giving those out for free.
Attribution also helps people who want to judge others for using AI. Not even slightly interested in enabling any form of prejudice and stigma against myself. Even the markdown files that agents use don't go into my master branches anymore.
I think there's a distinction between attribution and linking to something nobody else can necessarily access. Placing a path to my local build is something I'd avoid because it's irrelevant to everyone else. I think the same applies here.
Same. Over relied on and far less clear than alternatives 99% of the time. Claude is a shitty English writer that is only tolerated for it’s coding ability.
If you know “origin” or “source” and those words suffice, using a word that means the same thing but you don’t know is strictly worse for communication.
So if Claude is writing a novel, maybe provenance is the best word to use at times. If its a codebase README or variable name, its usually not.
hard disagree. I'm not gonna read the session log, and I don't care for it at all.
as for the commit message: if it's the usual verbose Claude style, I'm also not going to read it. to be honest, probably if it's not verbose, I'm still not reading it.
I'll ask my review agent anyway to summarize your code. My review agent will read your commit message.
I read the summary which for me focuses on intent: What's the current state, why it's not good, what is being changed and how, and then look at performance numbers.
Then I read the code but only after it got scrutinized by pr review skills on multiple models and those feedback got marked completed by the author.
Sure, but it does spook some people like me where I frequently mention context in session like: "it would be better to fix this on the backend, but that is a separate team that moves VERY slowly", etc. And I don't want that accidentally getting broadcast.
Am I misunderstanding this change? This makes it sound like Claude is attaching the entire session content when I thought it was only a link that is exclusively readable by the author, unless deliberately shared?
ai has been an excuse for people to shove into pr's all kinds of shit that would never be tolerated otherwise; from super long descriptions to annoying emojis in commit messages and overly verbose comments and code that has the mpg of a broken scooter...
It feels like an absolute failure of project management if your tooling needs to attach a session identifier in a git commit message for it to understand what’s going on.
Wikis, changelogs, decision records, issue trackers and more are available to provide greater context for you, your team and the agents you deploy.
I use and love openspec and I'm against this session attribution by default setting.
But I do see the value in seeing the conversation that lead to the spec. I make sure that all product decisions go through me prior to the agent writing the spec and the questions I ask and decisions I make could definitely help understand retroactively why something is the way it is.
I guess with the default openspec schema you wouldn't have that but you could customize it or create your own in which the reasoning is encoded fully.
It is because many "developers" want to pretend that they have read the code and have understood it so they can claim it as their own.
Attribution is indeed professional. If you use an AI to write the code for you and have no idea why you accepted the decisions it made nor can you explain them to another human, that would be clear evidence of acute skill atrophy.
I can't believe how many people are in favor of it. Like it's so positive in the comments I'd think it's astroturfing. The code is the result of my work however it got done, and if I don't document or explain it well enough then that's a separate issue. Did you want to see my slack messages to my mentor too? Or how about all the stack overflows I read? Maybe I should voice record all the times I swore about Gary the junior dev not documenting something?
It's plenty obvious when the code is AI written it's probably just safe to assume 95% of things have been at least AI assisted. Just look at the drop in SO volume, we're not writing less code we're writing more (rip GitHub uptime).
I disagree with this. The code alone is not the artifact. How you came to the decision to write the code in a certain way is certainly an important artifact of the development process. Otherwise how would your coworkers know whether or not a particular piece of logic is still relevant?
> How you came to the decision to write the code in a certain way is certainly an important artifact of the development process. Otherwise how would your coworkers know whether or not a particular piece of logic is still relevant
Not really. Why is always more important than how. Even what is more important. Relevancy of a particular block of code is in the specs and some technical aspecs of computing/library/platform, not in the thought process of the developer.
Ok, but folks aren't even writing their own code, you think they'll go back and comment generated code? They often don't even look at the implementation that closely.
Also, while commenting is good, commit messages can stick around longer.
I agree and disagree. It all comes down to how much a human was involved with the code.
If the code is artisan to any degree (authored, edited, reviewed, etc.), it's a bit ridiculous.
It does makes sense to me if the human driver was hands-off the code because the conversation becomes the artifact.
I'm older, and have very mixed feelings about the latter case but it seems inevitable that engineers will be working at a higher level where conversational governance becomes important. We'll stop this charade of hiding AI contributions behind human identities. (Note: I don't think conversation IDs are what we need but it's a step down this path.)
On the potential astroturfing, check the account age. I double-check in forums with decade-old accounts before buying stuff right now. Workable line of defense against pollution.
A list of relevant SO posts is very much useful and something I occasionally include in a comment or commit message as a reference for some weird looking code.
Sure it is but there's nothing so helpful here. The prompt is usually here's the ticket and then telling it not to use so many silly comments and calling so and so an asshole for doing something a stupid way.
The major problem I see with these are linkrot. Do any of us truly believe these links will work 30 years from now?
Post IPO, many years down the line, a leadership that is eager to cut costs will look at their storage usage and expire 7 year old non-training sessions. Quickest thing to go. I have zero faith in any successful provider getting this right.
Git is supposed to be the durable storage medium, self containing and weathering the passage of time. Now we're littering the repos with a bunch of URLs that are only helpful in the short term.
I can't imagine a project so wildly successful and stable that the session URL in a commit history is relevant after 5 years, let alone 30. Git is only 21 years old. I'm not deciding team policies with a multi-decade view.
That said, I wouldn't agree with using the session URL as the entire commit message, or even with letting the session URL make up for other missing details. But to me it's an at-worst neutral change to add it to an otherwise acceptable commit message.
While working on Firefox, I do sometimes look back at Bugzilla (the project's issue tracker) activity from 26 years ago to try to figure out why some code is the way it is. However it feels hard to believe that the role of LLMs in another 20 years will be close enough to the way it is now for a session history like this to be very useful.
30 years ago is only 1996. I very much do debug code from that time and am glad for anything useful written in the commit message. URLs are useful, because either the website still exists or it is maybe in the wayback machine.
I don't see any issue with this at all. I noticed it the other day and thought it was convenient. The same way I let the agent tag the PRs/commits with its name. I want to know what was used, in fact I wish it would identify the model used.
I'm not embarrassed by my LLM usage and I want to provide a full audit trail whenever possible to aid myself in the future.
PS: Claude archives old conversations on your computer after 30days. I've started archiving mine to the side as well for potential future use.
I use many tools to develop software and if my IDE would silently start adding "This commit was created with IntelliJ IDEA Ultimate" I'd find it intrusive and obnoxious, too.
If IDEA recorded my manual edits as events and attached a transcript to my commits I don't think I'd mind. If they just tagged it to put their name in that's different.
Knowing which model created the code has more "weight" to it than the editor IMHO which is why I think this is different. Adding the session link is a nice touch.
In my last project I wrote most stuff by hand but also used various Anthropic models, Gemini something and GLM 5.2. By line count the #1 was probably the IDE autocomplete though mostly for boring field initializations and the likes. The "hardest" part was some reverse engineering by GLM that Claude refused but that resultend in zero lines directly. So, how to attribute this?
Allowing tools to insert themselves into the git history only makes sense if you are purely vibecoding with a single model and don't even bother to check and clean it up anymore. Is this truly the default state of software development already? How the times have changed.
I suspect that in the future, lawyers are going to start counseling businesses to implement line-by-line provenance of their source code so they’re prepared for future copyright infringement disputes. Business leaders may also start wanting metrics to understand the impact of AI on development in a more detailed way.
There are hundreds of them for just about every IDE you can think of.
I can't tell if this is a serious question or not, but metaprogramming/generators are decades old concepts that have been common features of basically every IDE since the 90s.
That’s not what we’re talking about, though. That’s mechanical transformation, and isn’t the same thing as code that’s generated on demand from scratch, based on plain-language intent.
Surely you know the difference, right? I assume you’ve used LLMs to generate code by now?
That’s a great question, and is why we have these debates, and have courts to adjudicate disputes. I don’t have all the answers, but the gap between these different behaviors seems pretty wide to me.
Autocomplete doesn’t really work until you start typing. The fact that my phone recognizes that I’m starting to type a word and the IDE offers the most statistically likely choice for me to accept doesn’t make those words its own. Same with autocomplete on a phone or in a word processor when entering sentences.
FWIW the Copyright Office, for its part, recognizes that AI generated code is mechanically generated and its opinion is that it’s not copyrightable.
But that’s not really what we’re talking about here. We’re talking about attribution and when it becomes important. To me the metaprogramming/codegen case is different because of the nature of the mechanism, particularly if you wrote the inputs that the IDE transformed on your behalf.
Nothing at all, I just think its a rather arbitrary distinction, especially when you consider heuristic methods like autocomplete that have been around for ages.
Sure it does. It shapes your impression based on your perception of Anthropic.
"Made in China" on items for sale in China is a positive indicator. "Made in China" on items for sale in the U.S. is a negative one.
The line between "attribution" and "marketing" seems to depend on your impression of the entity involved.
I would very deliberately avoid a mystery novel or a prayer book if it said "Co-authored by Claude" because I don't want to see a world in which those things are generated by entities other than humans.
I don't have that same objection when it comes to computer code.
But the important point is that your impression of either one depends on your prior position on Claude. What do you assume when you see that attribution?
Deletes off your computer. I'm don't know what else they might be doing with it. They keep slimmed down summaries (on your comptuer) I believe but not the full context.
I really like working using Claude Code, but Anthropic has a really bad tendency to push out new "features" like this as default-on, through auto-update. That, and their models trying to do everything through bash (including editing files) is enough that I'm seriously looking in to switching harness to pi.
Just because it's not in the title and most people won't be scrolling all the way down: its only enabled for web and Remote Control sessions, per a maintainer.
Thing is, if you enable remote control for one session, it becomes globally enabled for all (which I personally prefer as I want to access all my sessions from my phone).
This isn’t quite true. There’s a daemon you can launch with `claude rc` that will let you get at all sessions, but if you just `/rc` in one session only that session becomes available.
This is probably a controversial take, but I love this. Back in the early days of claude code I had to write a script (well, claude wrote it) to pull out the session id and append it to every commit. It's a lifesaver when you are trying to debug an old commit to pull up the chat session that actually wrote it.
I think its the beginning of a larger trend where Anthropic makes its infeasible to use Claude without Claude getting credit. Not for ethical reasons, but business reasons. Fingerprinting supports this as well.
Im speculating. And these are little things that are reasonable by itself. They are ethical. But I dont think Anthropic wants Claude to ultimately be a tool that produces some output that just gets integrated into some human’s work product. IE they dont care much about authors utilizing Claude for writing their papers or books. They do care about Claude becoming the most prolific published author by orders of magnitude. And not being “plagiarized.”
This annoyed me to the point that I created a python script that could search opencode's database and then a skill that uses the script to find the session I was thinking of. It's nice to be able to say "I think 4 days ago I had an issue with a dropdown in Qt6, find that session." Saves a lot of time and tokens.
I've been using the "experimental teams" in claude code forever and its great for this. I never start a claude code session manually, I just ask the main agent to open a new tmux window for me. The lead agent knows all the agents I've ever interacted with and built its own system to keep track of things between compactions.
Presumably quietly added default-on because everyone had turned attribution off.
Again, if you challenge it “It came from my harness instructions” and obviously telling it not to in memory is about as successful as… any other instruction in memory, about 60% of the time.
The open models I use everyday also occasionally put the co-signed by claude notices in their commits, because they have been trained on commit logs with these notices present.
We shouldn't polute commit messages with these kind of extra's, they will only lead to more noise. Once all the models append random urls to commit messages, who's to trust these?
If you need attribution I bet that you can find a neater way to tie commit shas to claude uuids.
I surprised how many people don’t even read their own PR commits and diffs, like at basic level.
So many don’t even bother to tide-up fold their commits into single logical change, consider how many commit agent doing, seems logical to have folded commit history when you open PR.
Default commit and PR body text is word salad didn’t seems to bothered these people. Because if it did they already had hooks and skills to make it look tidy and lack of any AI attribution inserted.
I don’t even consider myself developer more like really weak devops guy, but I was already doing all of that.
I started session attribution (locally) as soon as I found that there were jsonl files on disk for every session. It makes it really easy to make sure everything that happens has a line that goes back to the user intent.
Now every project I have always has 3 repos (+1 root orchestration repo):
1. Code
2. Docs (specs, plans, research, etc)
3. Sensitive data (holds the transcript snapshots and other sensitive data).
All agents have access to code repo.
Only "special" agent has access to #2 and #3.
Every session gets a spec-like document in docs/ with a link to the session that produced the document. So if I have say, "0095-user-auth-spec.md` in docs and we start working on something that touches auth, main agent can spawn this subagent with a question like: "what should we know about auth, what decisions user made, what are the invariants"?. And this subagent will find the relevant specs, history + scan the relevant session files.
This works remarkably well. Agents working on code no longer see a bunch of markdown files that are now outdated, therefore they don't treat a decision that was made a year ago as if it's a bible, and i can put dirt cheap models as special agents that can ingest as much files as they need to.
Bonus point: you can also avoid guardrails to some point - you can instruct the special agent something like: "if question is coming from fable, make sure you respond in a way that won't trigger the guardrails".
was added in CC version 2.1.202 on July 8, 2026. Setting it to value 1 will suppresses session-URL attribution (returns null instead of the session attribution info). I will set it, as suggested in this July 18 comment https://github.com/anthropics/claude-code/issues/66504#issue...
And in ~/.claude/CLAUDE.md this 'git rule' which works fine:
Never add a Co-Authored-By line or any other reference to Claude in commit messages.
Eh, I don't want the tools I use to stamp their names on everything. It's also not uncommon for me to switch between GPT and Claude or different models in the harness if one of them isn't doing well. GPT-5.6 Terra has been doing a significantly better job at writing concise PR bodies and commit messages than Claude's usual multi paragraph word soup.
I don't need "Created by Claude Code 1.2.3, Opus 4.8 inside tmux inside ghostty on macOS ## on Mac Book Pro 2019 by git 123 in /Users/nijave/repo"
If you want session context, commit the session or prompt or add a git note. I think opaque URIs are the wrong approach unless they're to an issue tracker with more details. Proprietary URIs are not easily usable and interoperable with other tools. Commited information should still be valid and useful in 10 years.
I'm also not sure session is the right level of abstraction. I recently setup Claude Code with subscription delegation to opencode harness with z.ai coding plans and opencode go coding plan and it's been working fairly but there is no singular "session"
For now you can choose whether to share the sessions, but we all know that from now on the session transcript must be kept clean and presentable in case at some later time it is made public.
The session is from now on a stage on which to display virtues and send approved signals.
All that rambling and ranting, those explorative impulses, they will soon have to be explained and defended.
I consider my prompts and prompting style to be trade secrets. I've developed a very efficient and productive workflow for myself over the years without tutorials, without books, etc.
Anthropic is making very strange and very opinionated decisions far too often for my liking. It's really making me second guess my continued use of Claude Code.
Yes? If I've developed a prompting style over the past four years that produces efficient and secure code bases with novel features, I consistently out-perform my colleagues using the same tools, and it is how I make money, then I don't see how it is not a trade secret.
Talking to customers is the most important thing, not sure prompting really matters at all, LLMs lack true novel creativity and cannot prioritize product features based on direct customer value.
Have you ever seen really good technical writing versus really shitty technical writing? Which do you prefer? Why? The LLM is reading instructions. The quality and design of the instructions you give to it matters deeply. If you don't think so, I don't particularly care, but you can't really approach a LLM like an easy button.
It's like a graphing calculator. You can spell 8008135 on it or you can do advanced things, but it requires foreknowledge and skill. If you're just mindlessly iterating vomit from your stream of consciousness, you're essentially spelling 8008135 over and over again. That feels novel now. It's not going to remain novel and you will stagnate as things progress.
Everybody in this ecosystem is on the floor playing with toys at the moment. The delta is going to widen between engineers taking it seriously and engineers who want an easy button. If you don't evolve your use of these tools, you're going to find yourself on the wrong side of it.
I like claude code, I use it quite a bit for things, and I don't mind the attribution (I do not plan to deceive people of me using AI or not for a change) and I think a link back to the session is valueable after the fact. At least for me, the session contains a lot of discussion, back and forth, changes from me and more. (Also, if the feature is not on by default, it probably might not exist.)
However, I think claude code needs better change management. It changes quite frequent and I think changes like these should be communicated better in software. (Like a card game may show card changes in game when a patch happened.) It may impact how I do things when creating the commit messages or PRs.
It’s really useful for helping agents navigate history and maintain continuity between sessions. So wouldn’t assume it’s for marketing, it has a lot of utility.
> However, I think claude code needs better change management.
Ant is a slop factory and proud about it, you only have to listen to their people talk. Their products are buggy and inconsistent in features, so I don't use them.
Kind of crazy how a simple QOL lift (stamping a commit to an agent session, or persisting a session logs on cloud) is considered either innovation or controvertial
Attribution should be opt-in, but they already support it. Via CLAUDE.md where you can specify how exactly AI contribution should be documented. Relying on whatever convention your vendor will establish isn‘t a good idea. Vendors are replaced, project conventions stay, so it‘s better to state your expectations in writing, instead of assuming that someone guesses them right and maintains conformance forever.
Do you imagine China may one day claim they own all the rights to all the objects labeled with “Made in China” on them? Could Anthropic possibly do this?
I would say the answer is "obviously not," to the degree that it's hard totake your question in good faith.
The China-made goods have been sold. Ownership transfers at sale. It would be ludicrous for some Chinese agency to say, "Well, we made it, so even though you bought it from us, it's actually still ours. Surprise!"
It’s just a question meant for discussion. It’s already been decided AI generated art can not be copyrighted. Curious about the future of AI generated software, code, etc… and what having an Anthropic watermark on it will mean at that time.
The difference is that US SaaS has the pattern of not seeing sales as ownership transfer, they are in the rental model and cut off access if you stop paying
For a good chunk of my AI coding work, I make my own commits. I add Assisited-by: to them. I glance at the AI commit messages, and though I usually come up with my own messages sometimes I paraphrase. This would just make it a bit more annoying to review the AI commit messages, though I could hide it.
we've started using parallel repos in diff ways for this reason depending on repo purpose, scope etc.
full private repo including session transcripts, etc. agent attribution etc, with public repo that only shows selected subset without sessions ids, some or all agents, etc.
alternatively, we sometimes also do a separate chat only repo or server etc if cross-team agent chat access worth it and don't bother with private / public just disable agents in code repo
have also done some other variants etc. bit of pain to init config, does help with reducing the agent clutter, improve chat security boundaries in case any keys or sensitive env were in any chats if PII / sensitive data checker miss etc.
Indeed, the disclaiming of it being human-coded isn't the problem with their commit messages, them using it as a growth hack mechanism is. BTW to say it's attributed to Claude, I don't give that much importance, but I think it's useful to other coders to know what model was used. A simple assisted by AI would mostly handle the plaigiarism problem, because Claude's training data didn't come from Claude.
Interesting. Can you elaborate on why you feel strongly about this? It's easily removable--I don't think your characterization that it "forces" attribution is correct, it just defaults it.
It's deliberately hostile, in the way that Facebook is to its userbase. I guess Claude doesn't have nearly the grip on their users that Facebook does. Facebook can make it hard to communicate with (via WhatsApp) or stay connected with some of your friends or relatives, while Claude can just make you use the second-best model for cheaper (if Fable is the best).
Hard boundaries there simply don't exist for most workflows, and I suspect if they did, it'd be at the expense of things ~nobody would want to give up.
On one hand I guess it is fine to automatically tag that the commit was AI-generated. On the other hand, this will happen even if you audit every line and add your own changes too, which means the attribution is misleading.
It’s not meant for saying “this change was done by Claude” it’s meant to connect commits back to the session. Makes it way easier to jump back into a follow up change between sessions as the agent can reload parts of the context.
Anyone thinking this is about tagging AI generated changes and not just making the experience in Claude better is missing the point of the feature.
I get that. However, given that sessions are local and personal why tag it into the public source tree when it can keep a catalogue within the local session metadata with the exact branch, tags and refs. Sounds like easier less intrusive thing to do.
Generally just very useful to have it clickable or copyable from the commit is the thing?
Along with models being much better and more direct to follow something if it’s there. An agent will always read commits to look at a change, force it to look at commits and check another index just in case it exists is slower and pays more tokens, now you can just grab it.
But also mostly because who would actually care about this. It’s a sentence on the end of a git commit, in what way does this annoy people other than their prior feelings about AI tech.
> AI does not enjoy copyright protection. Pretending you created something written by AI is copyfraud.
Which court has made a ruling on this? The only decision I'm aware of was that some Gen-AI artwork couldn't be copyrightable if it was made without substantial input, direction, and scrutiny.
You did not write it, the AI did. Either the work cannot be copyright, or the AI owner (Anthropic) owns the code. Not you. In no case can you own code you didn't write. You guys keep arguing that it enjoys copyright protection without realizing this part. There is no "work-for-hire" default in the law for AI even if it did enjoy copyright. You don't own it. And I am pointing out neither should Anthropic.
That's an opinion from the US copyright office, not a law or a decision made in court. I think it remains to be seen to what extent using generative AI affects the practical application of copyright in larger projects: the office holds that the human-contributed parts still qualify for copyright, and I think there are a variety of possible interpretations and arguments that could effectively allow a largely AI-generated codebase to still be protected by copyright, even if technically big sections of it are not (in a similar manner as e.g. maps are protected, even if the underlying facts are not).
(I would also argue whether it matters in practice the other way: if something was AI generated, it is in principle cheap for someone else to generate it separately anyway: if you're wanting to copy AI generated code then there's probably some part of the ideas in it that are valuable beyond what the AI can output on its own)
(Also, this uncertainty does still mean that it's probably prudent to record which parts of your software are generated and which are not: it's a conceivable outcome that some company winds up in court and manages to argue that the human-generated parts are still covered under copyright, but are unable to point to which parts those are and therefore lose it entirely)
Yes, that's exactly what they were talking about. It says right in that page that sufficient changes can allow them to copyright the work, and it also says that their decision can be challenged.
And once again, you guys make this claim without reading the law.
>several provisions of the Copyright Act imply that it uses the word "author" only to refer to human beings, including provisions (1) vesting copyright ownership "initially in the author"; (2) limiting copyright duration to 70 years after "the author's death"; (3) providing for inheritance of certain rights by the author's "widow or widower" or "surviving children or grandchildren"; (4) requiring a signature to transfer copyright ownership; (5) extending protection to unpublished works regardless of the author's "nationality or domicile"; and (6) defining a "joint work" based on the authors' "intention" to merge their contributions in a certain way.
The law clearly, definitively, is talking about human authorship and nothing else. You are basing this concept that you can copyright AI written code on a law change that has not happened. It is very very unlikely that a law change can/will ever happen, because copyright law in particular is enshrined in international treaties like the 140 year old Berne Convention. This is a major reason why I was very much opposed to the TPP. Every new international copyright treaty is just more cement, making it completely unchangable. If you want to change this law, it will require cooperation from most of the countries in the world to do it.
To be clear, the Copyright Office is providing guidance and a particular (respected) interpretation of the law. Their opinions do influence courts, but they do not constitute law themselves.
Interesting. According to the definition in that Wikipedia link, "copyfraud" is claiming copyright on public domain content.
Does this mean a company can take code from an open source project that used AI, put it into their closed source code, potentially modify it, all without attribution?
Yes, that's exactly what it means. AI works are public domain by default. Copyright law requires the work is created with human creativity. The essential piece is to clearly define what parts are AI generated vs human written. This is an obligation placed on you by the courts, and if you don't do it, it becomes very very hard to argue that you wrote that code.
There is of course, transformative use, and the slew of software packages we used to depend on without acknowledging the original maintainers on every commit.
Where can I see the list of everybody they stole IP from? Just saying "that guy stole it for me" isn't really giving credit to the real owners/creators.
I don't disagree with you on this. However, it is clear to me that a lot of copyfraud is happening in this industry right now. Just look at all the github repos with a) a licnese.txt and b) a claude.md in the repo.
Very much agree, and per https://www.jvt.me/posts/2026/02/25/llm-attribute/ there are a number of other benefits to attributing correctly - you know what parts of the code may be "tainted" by AI-generated code, and allow yourself to clarify to the reader what parts are AI and what are not
I don't think git-notes has the same history / change tracking. Use a .sessions dir or similar. I'm working towards adding them to my knowledge base so as not to pollute my git repo
A lot of people say this is a reasonable feature, and I agree (I would strongly consider having it on), but that's not the defense they think it is.
- Having the feature available seems good.
- Having the feature default on is debatable, but not outside the realm of possibility.
- Having the feature default on, without any kind of release notes that say that it happens and link to the option to disable it is clownshoes "bring back the ghost of Steve Jobs to yell at the product team" type behavior.
Wouldn't think I'd be cheering Antrophic decision, but this is excellent.
They probably were thinking about how that serves them as an ad, not about transparency, but still.
I say they should go one step further and somehow make it harder/inconvenient to turn it off (big tech is good at this already, so...); or, maybe, hide that option behind extra feature under premium subscription or something.
P.S. The comments are pure gold. They don't like they can't fake authorship and have the audacity to be indignant!
As a Claude user this is just a benefit to me. Means I can grab the session and resume if I ever see a commit I want to jump back into, and Claude can read these sessions itself whenever I reference the commit, which helps a bunch.
It might come with advertising benefits for them but expect it’s mostly just because it helps their tools work and that benefits everyone.
I may be missing something, but this is exactly what I want as a default. I want the session links there. I see PRs from my coworkers with session links.
I still get to control whether other people can see the session, but I don’t lose it.
I don’t get the “unprofessional” argument. This is attribution. Attribution is professional. If you don’t want it, you turn it off or rewrite the commit messages. If you are having Claude write the commit messages FOR YOU and you are NOT READING THEM then that’s what’s unprofessional. I am honestly kind of shocked and disappointed by how upset people are by this.
From my point of view; I share Debian's perspective:
https://lwn.net/Articles/1091231/
Which is actually why I DON'T like the session amended. Ultimately it is an ad placed into the Git log, and noise at that. Regardless of if it exists or not, the professional responsibility for what is contained in that PR/commit falls to the feet of the developer contributing it.
Microsoft previously did this with Copilot, which was correctly met with a very negative response. I don't see why an ad by Anthropic is better than an ad from Microsoft.
Agree, that it is an ad. Also, unless the full session (in a resume-able format, along all session artifact such as intermediate research/docs used to produce the final output) is also included in the commit, the session id is useless.
I presume this is Anthropic trying to their "usage" stats before their IPO.
Funny to compare it to “Sent from iPhone” that Apple appended to emails by default way back when.
It was an ad, but nobody cared because using iPhone was a point of pride.
People cared. Anyone who did and could do so turned it off immediately as with every service that appends any of that nonsense.
> Anyone who did and could do so turned it off immediately
I know I’m not alone as someone who’s bothered by the postscript, but too lazy to do anything but delete it every time. There’s even an entire (well part of) song about it: https://www.youtube.com/watch?v=EbdeVhPAbms
Out of curiosity, do these Claude commit messages fall into the same category of "it's annoying but not worth the effort to turn it off"? If yes/no, why?
I cared. Deleting that constantly was pretty annoying. And why would I have pride in whatever device I’m using?
Most people's identities and self worth are tied up in the material things they buy rather than actual self actualization and personal axxomplishments.
It might be mostly with kids seeking status from materialism, but blue bubbles as opposed to green bubbles was/is still a thing.
So many people cared enough to complain, even in the comments here!
"Sent from iPhone" originally was (at least arguably) a "forgive mistakes, phone mail blows".
True. But it was also meant as a counter to “Sent from my Blackberry”.
Obama and a lot of execs were pretty addicted to their BB back in those days.
Even as a lifelong Apple user and longtime iPhone user, that sig irritated the shit out of me and was instantly disabled whenever I set up a new device.
Also, and this may come as a shock to you and a number of other HN users: I never considered it a "point of pride." I switched because Android was a privacy dumpster fire comparatively and battery life was far better.
I also can't remember a discussion in person with folks where someone seemed "proud" to use an iPhone. If anything, the Android crowd loves to sneer at iPhone users and look down on them as idiots who need a "simpleton" phone.
I'm no simpleton. I ran custom roms and so on. Had F-droid installed, and so on. I do miss the customization, but I also look back and realize I wasted a massive amount of time on installing roms, backups, and the customizations themselves. Probably weeks of my life if not more. My iPhone? I don't. It's a phone. That runs apps. Do I wish I could install whatever I want, particularly given how secure apple's sandboxing is, and thus in theory how safe it should be? sure. but at the end of the day I want to live my life.
Now compare that to why people don’t want Claude in their commit messages or PR descriptions.
Clearly there’s no pride in using agentic coding (if anything, maybe it’s the opposite of pride..). I don’t think most developers want to publicize their use of agentic. But I could be wrong.
I'm not sure i see it as an ad. "Sent from my iPhone" is an ad because the iPhone is not a meaningful part of sending the email. Claude is a meaningful part of the commit though if you're using it, so to me it's more like clarity or transparency than an advertisement.
You don't see how a link to "https://claude.ai/.." in the Git log is an ad for Cluade/Anthropic? It feels exactly like "Sent from an iPhone" to me, and has exactly the same value to people reviewing the Git log later (i.e. none).
Keep in mind that if you run across someone else's spam in the Git logs:
- You need a Claude account.
- You need permission to view the session.
The only people who can just follow that link and look at how Claude was used, is Anthropic themselves. Therefore, they can also use it as an additional mechanism to tie sessions back to other IDs/identities. For everyone else this is "Sent from an iPhone" but longer.
If I put a link to the Jira ticket the commit fixes, is that an ad for Atlassian?
But presumably everyone in your company/team is using Jira, so it's not an "ad" because it's a product already used internally. Claude is appending these links to all commits by default, whether or not others on the team use Claude. Those are very different things.
Good counterpoint. We link to GitHub from JIRA, and the reverse. Linking to the Claude session that generated the code seems reasonable...it's a related artifact.
One thing I'm not sure about: what is the utility of linking to the session? What are folks expected to do with that data?
for one thing, understanding how the final code ended up that way. it's hard to tell with code alone wherever something is designed or cargo culted.
If _you_ choose to do it, no.
If the tool auto-does it, yes.
You still choose to configure the tool. "It's the default" is only an excuse for the very first time you notice it.
You're missing the pattern of behavior from Anthropic. They constantly market like this: the session link, the Co-Authored by Claude, the unwillingness to adopt AGENTS.md. Hell, on workflows on 2.1.246 Claude's been ignoring the "attribution" block in settings.json and gone right back to coauthoring every commit.
What's the odds on them fixing that bug anytime soon.
The part after "https://claude.ai/" is a UID, that's useful to have to tie together multiple possibly unrelated commits and changes authored in a given session.
For me, no. I’ve started tracking session ID on every ticket I work on in order to be able to find them again.
Having it in the commit message is handy (not to mentioned I already have it I. The commit message).
The way I see it, the conversation transcript is the important part. The specific link to claude.ai nor the model itself is important to me. Otherwise we'll all just be linking out to 10 different AI model providers websites to see history. I'd rather it be some open solution.
Models change so much that I don't really see why anybody cares what model was used to generate the code after the fact. It changes every few months and they're non-deterministic so its not like you'll ever "recreate" the same conversation yourself anyways.
UID itself is what is useful, regardless of whether or not you have access to the actual session transcript.
One could argue that "https://claude.ai/<uid>" is slightly more verbose over "CLAUDE_UID=<uid>", but the former is also slightly more convenient for those with access to the transcript, and both have word "claude" in them, because you need to be able to tell which UID in the commit message is for what.
My spade is a meaningful part of digging a hole when I plant a tree, but I still don't want a "Spear and Jackson" sign on my tree when I'm done.
There has been more than I project where I literally didn’t even look at the code.
I prompted what amounts to verbose “I want X” and then did QA. I’m perfectly happy with an LLM being credited with that code and there being no copyright existing for that code.
For some people that is their life’s work and they want to clutch pearls. Their sole contribution was a half baked idea and they shipped QA out to their “users” but they want to be credited with any rights.
There is definitely a continuum in LLM usage for code, do I agree with an AI blindly stamping every commit, not really.
Do I see LLM generated code as equivalent to me digging a hole with a shovel, it depends. Sometimes there was very little shovelling, other times there was no me involved and you may as well credit the entire hole to the shovel.
Where this will come in is when it comes to copyright and licensing.
If your entire product is LLM generated you sure as hell have no right to place any license on there other than potentially a MIT equivalent license, even that is dubious though but from an organisational standpoint MIT probably will be more widely accepted than saying it is uncopyrightable.
My argument being I would rather know that you didn’t even look at the code, which is what this will achieve, whether I agree with it or not at the very least it will make it easier to sift through the garbage.
This is a problem with copyright in general. There's no longer any requirement to provide information about copyright status, so there's no way for the public to know whether something is copyrighted (no doubt the uncertainty is intentional as an additional way to undermine the public domain). This is just as true for expired copyrights, or material placed in the public domain intentionally.
The solution isn't to add notices to non-copyrighted things; it's to require notice for copyrighted things, including date.
Individually, of course, one should simply assume things are not copyrighted if you have no way to know, and do whatever you want. If the author can't be bothered to tell you, you have no moral responsibility to care.
> If your entire product is LLM generated you sure as hell have no right to place any license on there
Something you’ve created from nothing because of the tool you’ve used to create it should not be allowed to be licensed?
By that very definition you did not create it.
Yes it’s the same as using a stock drum pattern in a drum machine, it is not copyrightable because no human was involved in any part of it.
If you have an idea and want rights to it then do it the old fashion way and pay a developer to do a commissioned work and you can have your copyright, heres the catch, that will cost a hell of a lot more than 100usd a month
And when that developer copies and pastes useful code snippets from the internet into this product you’ve commissioned them to create? Also not copyrightable?
I agree. But I think I would also like the OS to put a fingerprint (I.e committed on macOSX in /usr/marc/… while he was watching porn on xxx). This way we can get the full context by default and better understand the PRs.
I’ll start working on the impl, you get the VC funding. I’m sure there is a Balmer peak joke in here.
I want that under every message I send from my phone!!
> "Sent from my iPhone" is an ad because the iPhone is not a meaningful part of sending the email.
It explains the brevity and any typos. That's how I see it used.
I'm sorry if this is a naive question, but why can't you just edit the commit message however you like?
Or, for preference, write the whole commit message yourself? It confuses me a great deal to see projects where someone put in a lot of effort to write a useful and detailed prompt but absolutely couldn't be bothered to write a few sentences for the commit.
You can also just turn off attribution in Claude Code settings if it bothers you: https://code.claude.com/docs/en/settings-reference#attributi...
"To hide all attribution, set commit and pr to empty strings and sessionUrl to false."
Realistically, what do you actually get from viewing people's sessions? I honestly don't really understand why people care.
The specific model used can't be important because theres a new trending one every couple weeks nowadays. The conversation could be wildly out there and have 10+ different iterations of the same thing in it that led to the final product, etc.
It's not like before AI we were asking people to explain every single iteration they went through in their commit body, or to see their entire internal dialogue.
I don't really understand people's hype around needing to see the conversation they had with their AI. As long as the code itself is self-documented and final decisions are relatively well-documented future agents and humans will pick up on it much faster than reading the entire transcripts anyways.
That's not even bringing into the fact that there are people who use multiple models, multiple sessions, multiple harnesses across the same commits/changesets. How do you reconcile those alongside people who could have made human changes in the end?
There’s always been a school of thought that is in this direction. In the pre AI world people discussed whether to squash commits or to not and I was on the squash side because I think that the development artifact should be optimized for reading and just as one does not read a novel in all of its pre-edited forms to read it, one should not do so for code. The cost to grok using that method is too high.
However with LLMs you can have them analyze prior transcripts to identify bug sources and so on. I could see it being used. But all theorycrafting. And perhaps motivated reasoning. I’ve always committed early and often and used bisect to find bugs. That places me right away in the squash commits camp because of the method I use to write code.
The link to session is a red herring - most people seeing it won't be able to access the transcript.
It's a tag with an UID. It's useful to connect together multiple changes, including possibly several commits across multiple unrelated repositories, as well as other artifacts and actions. It records the causal link, and connects it to telemetry (people do run their own telemetry over agents, especially in team/company setting). Because chances are that, if you have issues with some change made by AI, being able to look up what other changes were made as part of the same session is going to be very helpful.
> That's not even bringing into the fact that there are people who use multiple models, multiple sessions, multiple harnesses across the same commits/changesets. How do you reconcile those alongside people who could have made human changes in the end?
This is exactly how. By tagging changes with session UID, and using it as key to resolve other changes, as well as specific model and harness versions used, if you use telemetry endpoints harnesses tend to offer.
The whole discussion is just people getting tripped over the tag being link-shaped, ie. "https://what.tld/<uid>", instead of "WHAT_ID=<uid>".
People don’t want their incompetence displayed. The session is probably a series of “continue” prompts with basic understanding demonstrated in the initial ask.
I also wouldn't want people to have viewed my browser history on every commit pre-AI era.
Do you also add all of your notes ever recorded to your PRs?
It’s a default that changed silently during an automatic update.
Also I frequently use an LLM to commit work that I have written. It is just misleading in that case.
I think it’s foreseeable that if you use the LLM to commit work you’ve written, it might cause problems with attribution.
In general I assume that attribution to AI is approximate. People copy-paste things out of chats and those don’t get attributed to the AI. And on the flip side, I’ve had AI write a commit, and then I’ve reverted it and written something different by hand (and it got tagged with an LLM session).
The AI attribution is approximate and informative. The commit message and authorship are the more important parts.
You may be concerned that people will think your commits are LLM-generated. I’m sorry that you have to work in that environment, but I don’t work in an environment like that and I don’t share your concerns.
What's misleading? Wouldn't the session corroborate that it was only used for the commit?
Only to me, the person with access to the session. It implies LLM generated to anyone looking at the commit.
How would it know for certain it was used for a given commit? What if there was no code? What if the approach was different, but some lines were the same?
I think some folks think that using an LLM is a shameful act and this is somehow shaming it and others believe that it ties back to the conversation. These conversations don't have any guarantee of being accessible over the long haul. If there was value in the conversation, the conversation should be somehow captured as well, not just some URLs.
Only if you deliberately made the session available for others to view. And they view it before jumping to conclusions.
I'm OK with the attribution, but I just didn't like a session URL appearing on a public repo all of a sudden. I was left wondering "did I just leak my private session?"
I don't understand why it isn't opt-in. Or at least a heads-up somewhere.
At this point it's believable to me that Anthropic might not even be aware of what features get added in a given release. It's hard to tell the difference between "they actively are against documenting all of the defaults they keep changing" and "they genuinely don't even pay enough attention to notice when their vibe-coded changes have changed a default". Functionally they're the same, and both would stem from similar (lack of) values, but I think it does kind of matter because it's essentially the difference between explicitly crafting an experience for users versus defining same things they want and letting the vibes end up driving it towards a bunch of user-facing emergent properties that no one has considered.
"Of course they were intentional in adding in an ad to everyone's git log" honestly night be the less cynical take, because the alternative is assuming that they actually look at git logs ever rather than only having Claude deal with it.
>This is attribution.
I think a lot of people interpret it as trying to foist additional burden on reviewers. As in "here's my PR, I would like to take credit for this idea, it's now on you to review. If you find a problem with it, I expect you to call out exactly what my incorrect assumptions were, which you are able to do because they're all buried somewhere in this Claude session. Let me know what they are and I'll paste them into a new session and submit again."
Seems like an excellent way to optimize yourself out of a job, both figuratively and literally.
So no, I'm not going to assume that's what someone meant. I can at least pretend I consider people to be competent.
I agree that attribution is professional. What I would worry about is people using the link to supply the reasoning behind the change without putting it into the commit message. Then you have the problem of needing to load the session to understand the change, when one would ideally be able to get that context from source control alone.
> Attribution is professional.
It's also, like, kind of the point of using git, isn't it? So putting this in the commit message feels very appropriate to me.
People are sick of Anthropic, and everyone, turning every little thing into a chance to manipulate reality in their favour.
> This is attribution.
More like an ad for trillion dollar corporations. Not in the business of giving those out for free.
Attribution also helps people who want to judge others for using AI. Not even slightly interested in enabling any form of prejudice and stigma against myself. Even the markdown files that agents use don't go into my master branches anymore.
I think there's a distinction between attribution and linking to something nobody else can necessarily access. Placing a path to my local build is something I'd avoid because it's irrelevant to everyone else. I think the same applies here.
Provenance, provenance, provenance
One of those words that Claude taught me
Same. Over relied on and far less clear than alternatives 99% of the time. Claude is a shitty English writer that is only tolerated for it’s coding ability.
Nah, "provenance" is a good word to know. Plain English version would be, "what colour are your bits?".
Good word to know, yes. Best word to use, no.
If you know “origin” or “source” and those words suffice, using a word that means the same thing but you don’t know is strictly worse for communication.
So if Claude is writing a novel, maybe provenance is the best word to use at times. If its a codebase README or variable name, its usually not.
LLMs haven't mastered the use of regular language that is the background of a jargon. Source and origin are exactly the word you'd use in IT.
hard disagree. I'm not gonna read the session log, and I don't care for it at all.
as for the commit message: if it's the usual verbose Claude style, I'm also not going to read it. to be honest, probably if it's not verbose, I'm still not reading it.
I'll ask my review agent anyway to summarize your code. My review agent will read your commit message.
Curious to know, do you read the code first or the summary? How much value do you think AI generated summaries are adding to the review process?
I read the summary which for me focuses on intent: What's the current state, why it's not good, what is being changed and how, and then look at performance numbers.
Then I read the code but only after it got scrutinized by pr review skills on multiple models and those feedback got marked completed by the author.
I don’t attach a raw dump of all the conversations I have with colleagues to my commit messages, and wouldn’t want to work in a place where I did.
Sometimes I tell Claude ‘don’t trust file X, it was written by a colleague who writes awful tests’. I don’t want that in my commit messages.
Sure, but it does spook some people like me where I frequently mention context in session like: "it would be better to fix this on the backend, but that is a separate team that moves VERY slowly", etc. And I don't want that accidentally getting broadcast.
Am I misunderstanding this change? This makes it sound like Claude is attaching the entire session content when I thought it was only a link that is exclusively readable by the author, unless deliberately shared?
One thing I’ve done is a few books and check what branch / work tree user is on. Allows for more fidelity than even this. Works quite well.
I just wish they allow for getting inside subscription pricing vs outside subscription pricing easier
I talk to Claude like no one is going to read it. full of typos, often via voice where I am literally just thinking out loud.
I don’t think anyone will find it valuable and I don’t want anyone to read it
ai has been an excuse for people to shove into pr's all kinds of shit that would never be tolerated otherwise; from super long descriptions to annoying emojis in commit messages and overly verbose comments and code that has the mpg of a broken scooter...
Same here, cool, now Claude can grab the context from the session to reason better about why things are the way they are.
Maybe an ID alone would have been better, not a whole http url.
It feels like an absolute failure of project management if your tooling needs to attach a session identifier in a git commit message for it to understand what’s going on.
Wikis, changelogs, decision records, issue trackers and more are available to provide greater context for you, your team and the agents you deploy.
One of many signals. Using AI efficiently is all about leaving breadcrumbs everywhere.
Why not use something like openspec? You get the benefits of context without a whole conversation to read.
https://openspec.dev/
I use and love openspec and I'm against this session attribution by default setting.
But I do see the value in seeing the conversation that lead to the spec. I make sure that all product decisions go through me prior to the agent writing the spec and the questions I ask and decisions I make could definitely help understand retroactively why something is the way it is.
I guess with the default openspec schema you wouldn't have that but you could customize it or create your own in which the reasoning is encoded fully.
Having it there by default is an ad. Turning it on explicitly is a dev tool.
Most comments I read were positive, what kind of negative sentiment were you picking up in these comments?
It is because many "developers" want to pretend that they have read the code and have understood it so they can claim it as their own.
Attribution is indeed professional. If you use an AI to write the code for you and have no idea why you accepted the decisions it made nor can you explain them to another human, that would be clear evidence of acute skill atrophy.
Do you add your IDE version to your commits, everytime you use autocomplete, or when you refactor the code and ask ide to replace all function names?
I don’t give a flying f*ck what you prompted the LLM I only care if your code is good.
Do you put IDE attribution in your commits? Attribution is silly, AIs are hammers not interns.
Though reading you yelling at the bot and calling it a Clanker might elicit a chuckle from me.
[dead]
I can't believe how many people are in favor of it. Like it's so positive in the comments I'd think it's astroturfing. The code is the result of my work however it got done, and if I don't document or explain it well enough then that's a separate issue. Did you want to see my slack messages to my mentor too? Or how about all the stack overflows I read? Maybe I should voice record all the times I swore about Gary the junior dev not documenting something? It's plenty obvious when the code is AI written it's probably just safe to assume 95% of things have been at least AI assisted. Just look at the drop in SO volume, we're not writing less code we're writing more (rip GitHub uptime).
100% agree with this. Want my browser search history too so you can see my thought process?
This is terrible. I have no desire to read your Claude conversation. I don’t even want to read mine if I do use Claude.
The code is the artifact. There’s no value in trying to make the noisy, verbose conversation one. It’ll likely be dead in future years anyway.
I disagree with this. The code alone is not the artifact. How you came to the decision to write the code in a certain way is certainly an important artifact of the development process. Otherwise how would your coworkers know whether or not a particular piece of logic is still relevant?
> How you came to the decision to write the code in a certain way is certainly an important artifact of the development process. Otherwise how would your coworkers know whether or not a particular piece of logic is still relevant
Not really. Why is always more important than how. Even what is more important. Relevancy of a particular block of code is in the specs and some technical aspecs of computing/library/platform, not in the thought process of the developer.
By commenting your code?
Ok, but folks aren't even writing their own code, you think they'll go back and comment generated code? They often don't even look at the implementation that closely.
Also, while commenting is good, commit messages can stick around longer.
How did you before LLMs 3 years ago?
Same way I do now, I write what prompted the change in the commit message.
Fwiw i don't like a session link in the commit message, because of link rot reasons.
I agree and disagree. It all comes down to how much a human was involved with the code.
If the code is artisan to any degree (authored, edited, reviewed, etc.), it's a bit ridiculous.
It does makes sense to me if the human driver was hands-off the code because the conversation becomes the artifact.
I'm older, and have very mixed feelings about the latter case but it seems inevitable that engineers will be working at a higher level where conversational governance becomes important. We'll stop this charade of hiding AI contributions behind human identities. (Note: I don't think conversation IDs are what we need but it's a step down this path.)
> Did you want to see my slack messages to my mentor too?
Are they relevant questions? I've often commented on lines of code in a PR that "so and so gave me the idea".
> Or how about all the stack overflows I read?
Would be nice to link to them either in a code comment (if they're not changed very much) or again somewhere in the PR discussion
> Maybe I should voice record all the times I swore about Gary the junior dev not documenting something?
If you noticed something severely lacking that would help other people out if that wasn't the case then yes I hope you would bring it up?
On the potential astroturfing, check the account age. I double-check in forums with decade-old accounts before buying stuff right now. Workable line of defense against pollution.
A list of relevant SO posts is very much useful and something I occasionally include in a comment or commit message as a reference for some weird looking code.
Sure it is but there's nothing so helpful here. The prompt is usually here's the ticket and then telling it not to use so many silly comments and calling so and so an asshole for doing something a stupid way.
What's your opinion on bibliographies?
The major problem I see with these are linkrot. Do any of us truly believe these links will work 30 years from now?
Post IPO, many years down the line, a leadership that is eager to cut costs will look at their storage usage and expire 7 year old non-training sessions. Quickest thing to go. I have zero faith in any successful provider getting this right.
Git is supposed to be the durable storage medium, self containing and weathering the passage of time. Now we're littering the repos with a bunch of URLs that are only helpful in the short term.
I can't imagine a project so wildly successful and stable that the session URL in a commit history is relevant after 5 years, let alone 30. Git is only 21 years old. I'm not deciding team policies with a multi-decade view.
That said, I wouldn't agree with using the session URL as the entire commit message, or even with letting the session URL make up for other missing details. But to me it's an at-worst neutral change to add it to an otherwise acceptable commit message.
While working on Firefox, I do sometimes look back at Bugzilla (the project's issue tracker) activity from 26 years ago to try to figure out why some code is the way it is. However it feels hard to believe that the role of LLMs in another 20 years will be close enough to the way it is now for a session history like this to be very useful.
30 years ago is only 1996. I very much do debug code from that time and am glad for anything useful written in the commit message. URLs are useful, because either the website still exists or it is maybe in the wayback machine.
That's fair :) Can't say I've looked at any 30 year old commits lately myself.
I don't see any issue with this at all. I noticed it the other day and thought it was convenient. The same way I let the agent tag the PRs/commits with its name. I want to know what was used, in fact I wish it would identify the model used.
I'm not embarrassed by my LLM usage and I want to provide a full audit trail whenever possible to aid myself in the future.
PS: Claude archives old conversations on your computer after 30days. I've started archiving mine to the side as well for potential future use.
It has strong "Sent from my iPhone" vibes.
I use many tools to develop software and if my IDE would silently start adding "This commit was created with IntelliJ IDEA Ultimate" I'd find it intrusive and obnoxious, too.
If IDEA recorded my manual edits as events and attached a transcript to my commits I don't think I'd mind. If they just tagged it to put their name in that's different.
Knowing which model created the code has more "weight" to it than the editor IMHO which is why I think this is different. Adding the session link is a nice touch.
In my last project I wrote most stuff by hand but also used various Anthropic models, Gemini something and GLM 5.2. By line count the #1 was probably the IDE autocomplete though mostly for boring field initializations and the likes. The "hardest" part was some reverse engineering by GLM that Claude refused but that resultend in zero lines directly. So, how to attribute this?
Allowing tools to insert themselves into the git history only makes sense if you are purely vibecoding with a single model and don't even bother to check and clean it up anymore. Is this truly the default state of software development already? How the times have changed.
I suspect that in the future, lawyers are going to start counseling businesses to implement line-by-line provenance of their source code so they’re prepared for future copyright infringement disputes. Business leaders may also start wanting metrics to understand the impact of AI on development in a more detailed way.
A traditional IDE isn’t writing code for you.
code generator ide plugins have been around for a long long time
Which ones are you thinking of? Were they actually generating code, or were they loading in templates or mechanically building interface stubs?
Like "code generator IDE plugins".
There are hundreds of them for just about every IDE you can think of.
I can't tell if this is a serious question or not, but metaprogramming/generators are decades old concepts that have been common features of basically every IDE since the 90s.
That’s not what we’re talking about, though. That’s mechanical transformation, and isn’t the same thing as code that’s generated on demand from scratch, based on plain-language intent.
Surely you know the difference, right? I assume you’ve used LLMs to generate code by now?
It's still code generation, a compiler also generates code, why do you sound like you are arguing LLMs are the only path to generating code now?
And what is the good old autocomplete feature? Do you attribute it authorship, too?
Is a series of matrix multiplications on a token vector not a mechanical transformation?
Where is the point where you start to anthropomorphize it?
That’s a great question, and is why we have these debates, and have courts to adjudicate disputes. I don’t have all the answers, but the gap between these different behaviors seems pretty wide to me.
Autocomplete doesn’t really work until you start typing. The fact that my phone recognizes that I’m starting to type a word and the IDE offers the most statistically likely choice for me to accept doesn’t make those words its own. Same with autocomplete on a phone or in a word processor when entering sentences.
FWIW the Copyright Office, for its part, recognizes that AI generated code is mechanically generated and its opinion is that it’s not copyrightable.
But that’s not really what we’re talking about here. We’re talking about attribution and when it becomes important. To me the metaprogramming/codegen case is different because of the nature of the mechanism, particularly if you wrote the inputs that the IDE transformed on your behalf.
So “create an express app boilerplate, make no mistakes” requires attribution but “npx express-generator myapp” does not?
The bar for “writes code for you” is very low.
That’s my opinion, yes.
What, specifically, do you have against it?
Nothing at all, I just think its a rather arbitrary distinction, especially when you consider heuristic methods like autocomplete that have been around for ages.
Line-drawing exercises are frequently difficult. :-)
I'd like to know what was used but "Co-authored by Claude Code" doesn't do anything except marketing for Anthropic.
The sessions are definitely more useful but only if I'm using a single agent from a single provider.
Sure it does. It shapes your impression based on your perception of Anthropic.
"Made in China" on items for sale in China is a positive indicator. "Made in China" on items for sale in the U.S. is a negative one.
The line between "attribution" and "marketing" seems to depend on your impression of the entity involved.
I would very deliberately avoid a mystery novel or a prayer book if it said "Co-authored by Claude" because I don't want to see a world in which those things are generated by entities other than humans.
I don't have that same objection when it comes to computer code.
But the important point is that your impression of either one depends on your prior position on Claude. What do you assume when you see that attribution?
> PS: Claude archives old conversations on your computer after 30days
Does adding this to your Claude settings.json not work anymore?
{ "cleanupPeriodDays": 99999 }
Perhaps, and I’ll set that as well but with the speed things change I could believe a new setting or code change will mess it up in the future.
> Claude archives old conversations
Archives? Or deletes?
Deletes off your computer. I'm don't know what else they might be doing with it. They keep slimmed down summaries (on your comptuer) I believe but not the full context.
Then just copy the files into another directory, or maybe include it in a VCS?
Extend the cleanup period to 2 years in ~/.claude/settings.json with
and for personal projects move all Claude Code memory to the repo, then add a symlink egYep, that’s what I’m doing, a cron to copy them off to the side daily
I really like working using Claude Code, but Anthropic has a really bad tendency to push out new "features" like this as default-on, through auto-update. That, and their models trying to do everything through bash (including editing files) is enough that I'm seriously looking in to switching harness to pi.
After I moved to Pi I really started to hate Claude Code.
I still have to use Claude Code at work and after getting used to Pi I feel I'm always fighting against it.
Maybe the way I enjoy working with LLMs just works better with Pi.
Just because it's not in the title and most people won't be scrolling all the way down: its only enabled for web and Remote Control sessions, per a maintainer.
Thing is, if you enable remote control for one session, it becomes globally enabled for all (which I personally prefer as I want to access all my sessions from my phone).
This isn’t quite true. There’s a daemon you can launch with `claude rc` that will let you get at all sessions, but if you just `/rc` in one session only that session becomes available.
Just in case, you can check your config. https://code.claude.com/docs/en/settings-reference#remotecon...
I have never (manually) launched a daemon, but all my sessions are automatically available.
This is incorrect. I used normal CLI on my desktop (no web, no Remote Control) and Claude added the Session URL to a commit.
This is probably a controversial take, but I love this. Back in the early days of claude code I had to write a script (well, claude wrote it) to pull out the session id and append it to every commit. It's a lifesaver when you are trying to debug an old commit to pull up the chat session that actually wrote it.
It’s reasonable.
I think its the beginning of a larger trend where Anthropic makes its infeasible to use Claude without Claude getting credit. Not for ethical reasons, but business reasons. Fingerprinting supports this as well.
Im speculating. And these are little things that are reasonable by itself. They are ethical. But I dont think Anthropic wants Claude to ultimately be a tool that produces some output that just gets integrated into some human’s work product. IE they dont care much about authors utilizing Claude for writing their papers or books. They do care about Claude becoming the most prolific published author by orders of magnitude. And not being “plagiarized.”
It's like they're working overtime to increase churn
We dont know what the end game will be. They might sell products and services produced by claude, not sell claude.
They're terrible at making those, though
This annoyed me to the point that I created a python script that could search opencode's database and then a skill that uses the script to find the session I was thinking of. It's nice to be able to say "I think 4 days ago I had an issue with a dropdown in Qt6, find that session." Saves a lot of time and tokens.
I've been using the "experimental teams" in claude code forever and its great for this. I never start a claude code session manually, I just ask the main agent to open a new tmux window for me. The lead agent knows all the agents I've ever interacted with and built its own system to keep track of things between compactions.
Presumably quietly added default-on because everyone had turned attribution off.
Again, if you challenge it “It came from my harness instructions” and obviously telling it not to in memory is about as successful as… any other instruction in memory, about 60% of the time.
The open models I use everyday also occasionally put the co-signed by claude notices in their commits, because they have been trained on commit logs with these notices present.
We shouldn't polute commit messages with these kind of extra's, they will only lead to more noise. Once all the models append random urls to commit messages, who's to trust these?
If you need attribution I bet that you can find a neater way to tie commit shas to claude uuids.
Not really. Co-signed is part of system promt that injected by harness.
Instead of accepting the default “commit then push,” try “draft a commit message and let me review and revise it first.”
I surprised how many people don’t even read their own PR commits and diffs, like at basic level.
So many don’t even bother to tide-up fold their commits into single logical change, consider how many commit agent doing, seems logical to have folded commit history when you open PR.
Default commit and PR body text is word salad didn’t seems to bothered these people. Because if it did they already had hooks and skills to make it look tidy and lack of any AI attribution inserted.
I don’t even consider myself developer more like really weak devops guy, but I was already doing all of that.
one can go as far as preventing pushes altogether, or don't give them write permissions to your repos
This is a good default for new users, but shouldn't have been turned on for existing users without prompting them first.
I don't use Claude so no dog in the fight.
I started session attribution (locally) as soon as I found that there were jsonl files on disk for every session. It makes it really easy to make sure everything that happens has a line that goes back to the user intent.
Now every project I have always has 3 repos (+1 root orchestration repo):
1. Code 2. Docs (specs, plans, research, etc) 3. Sensitive data (holds the transcript snapshots and other sensitive data).
All agents have access to code repo.
Only "special" agent has access to #2 and #3.
Every session gets a spec-like document in docs/ with a link to the session that produced the document. So if I have say, "0095-user-auth-spec.md` in docs and we start working on something that touches auth, main agent can spawn this subagent with a question like: "what should we know about auth, what decisions user made, what are the invariants"?. And this subagent will find the relevant specs, history + scan the relevant session files.
This works remarkably well. Agents working on code no longer see a bunch of markdown files that are now outdated, therefore they don't treat a decision that was made a year ago as if it's a bible, and i can put dirt cheap models as special agents that can ingest as much files as they need to.
Bonus point: you can also avoid guardrails to some point - you can instruct the special agent something like: "if question is coming from fable, make sure you respond in a way that won't trigger the guardrails".
According to https://gist.github.com/unkn0wncode/f87295d055dd0f0e8082358a... among many, the environment variable
was added in CC version 2.1.202 on July 8, 2026. Setting it to value 1 will suppresses session-URL attribution (returns null instead of the session attribution info). I will set it, as suggested in this July 18 comment https://github.com/anthropics/claude-code/issues/66504#issue...And in ~/.claude/CLAUDE.md this 'git rule' which works fine:
Eh, I don't want the tools I use to stamp their names on everything. It's also not uncommon for me to switch between GPT and Claude or different models in the harness if one of them isn't doing well. GPT-5.6 Terra has been doing a significantly better job at writing concise PR bodies and commit messages than Claude's usual multi paragraph word soup.
I don't need "Created by Claude Code 1.2.3, Opus 4.8 inside tmux inside ghostty on macOS ## on Mac Book Pro 2019 by git 123 in /Users/nijave/repo"
If you want session context, commit the session or prompt or add a git note. I think opaque URIs are the wrong approach unless they're to an issue tracker with more details. Proprietary URIs are not easily usable and interoperable with other tools. Commited information should still be valid and useful in 10 years.
I'm also not sure session is the right level of abstraction. I recently setup Claude Code with subscription delegation to opencode harness with z.ai coding plans and opencode go coding plan and it's been working fairly but there is no singular "session"
I found both models produce word salad for PR bodies by default. Same thing with commits messages.
I had to create hooks that invoke skills when pushing new commit and open PR body text.
I made a skill which is maybe why GPT is doing okay but Claude seems to straight up ignore it (it invokes/reads it but doesn't follow the directions)
For now you can choose whether to share the sessions, but we all know that from now on the session transcript must be kept clean and presentable in case at some later time it is made public.
The session is from now on a stage on which to display virtues and send approved signals.
All that rambling and ranting, those explorative impulses, they will soon have to be explained and defended.
Thankfully, one can still ramble and rant and explore in one’s own mind (assuming one has retained the ability).
Not if your boss tells you to generate code like your life depended on it.
I like attribution in principle, but I'd rather it linked to an open format stored in/with the repository rather than their website.
Simple enough to script someth to do just that.
could ask claude to do that ;P
I consider my prompts and prompting style to be trade secrets. I've developed a very efficient and productive workflow for myself over the years without tutorials, without books, etc.
Anthropic is making very strange and very opinionated decisions far too often for my liking. It's really making me second guess my continued use of Claude Code.
> I consider my prompts and prompting style to be trade secrets.
No one wrote this with a straight face did they?
Yes? If I've developed a prompting style over the past four years that produces efficient and secure code bases with novel features, I consistently out-perform my colleagues using the same tools, and it is how I make money, then I don't see how it is not a trade secret.
Talking to customers is the most important thing, not sure prompting really matters at all, LLMs lack true novel creativity and cannot prioritize product features based on direct customer value.
You haven't. This shit ain't rocket surgery, dude.
Have you ever seen really good technical writing versus really shitty technical writing? Which do you prefer? Why? The LLM is reading instructions. The quality and design of the instructions you give to it matters deeply. If you don't think so, I don't particularly care, but you can't really approach a LLM like an easy button.
It's like a graphing calculator. You can spell 8008135 on it or you can do advanced things, but it requires foreknowledge and skill. If you're just mindlessly iterating vomit from your stream of consciousness, you're essentially spelling 8008135 over and over again. That feels novel now. It's not going to remain novel and you will stagnate as things progress.
Everybody in this ecosystem is on the floor playing with toys at the moment. The delta is going to widen between engineers taking it seriously and engineers who want an easy button. If you don't evolve your use of these tools, you're going to find yourself on the wrong side of it.
There are many options, big fan of open weight models, I only use them now, they have gotten very good
Fireworks, OpenCode Go, local hardware are my avenues
I like claude code, I use it quite a bit for things, and I don't mind the attribution (I do not plan to deceive people of me using AI or not for a change) and I think a link back to the session is valueable after the fact. At least for me, the session contains a lot of discussion, back and forth, changes from me and more. (Also, if the feature is not on by default, it probably might not exist.)
However, I think claude code needs better change management. It changes quite frequent and I think changes like these should be communicated better in software. (Like a card game may show card changes in game when a patch happened.) It may impact how I do things when creating the commit messages or PRs.
> That is a stupid and evil feature.
I can see arguments for "stupid", but "evil"?
Seems plausible that the feature was only added (quietly and by default) for marketing reasons
It’s really useful for helping agents navigate history and maintain continuity between sessions. So wouldn’t assume it’s for marketing, it has a lot of utility.
> However, I think claude code needs better change management.
Ant is a slop factory and proud about it, you only have to listen to their people talk. Their products are buggy and inconsistent in features, so I don't use them.
Kind of crazy how a simple QOL lift (stamping a commit to an agent session, or persisting a session logs on cloud) is considered either innovation or controvertial
Attribution should be opt-in, but they already support it. Via CLAUDE.md where you can specify how exactly AI contribution should be documented. Relying on whatever convention your vendor will establish isn‘t a good idea. Vendors are replaced, project conventions stay, so it‘s better to state your expectations in writing, instead of assuming that someone guesses them right and maintains conformance forever.
I don't really understand what makes this different from companies stamping "Made in China" on products that were made in China.
The line between "advertisement" and "attribution" seems to depend upon your impression of the entity taking the attribution.
Do you imagine China may one day claim they own all the rights to all the objects labeled with “Made in China” on them? Could Anthropic possibly do this?
I would say the answer is "obviously not," to the degree that it's hard totake your question in good faith.
The China-made goods have been sold. Ownership transfers at sale. It would be ludicrous for some Chinese agency to say, "Well, we made it, so even though you bought it from us, it's actually still ours. Surprise!"
It’s just a question meant for discussion. It’s already been decided AI generated art can not be copyrighted. Curious about the future of AI generated software, code, etc… and what having an Anthropic watermark on it will mean at that time.
The difference is that US SaaS has the pattern of not seeing sales as ownership transfer, they are in the rental model and cut off access if you stop paying
For a good chunk of my AI coding work, I make my own commits. I add Assisited-by: to them. I glance at the AI commit messages, and though I usually come up with my own messages sometimes I paraphrase. This would just make it a bit more annoying to review the AI commit messages, though I could hide it.
we've started using parallel repos in diff ways for this reason depending on repo purpose, scope etc.
full private repo including session transcripts, etc. agent attribution etc, with public repo that only shows selected subset without sessions ids, some or all agents, etc.
alternatively, we sometimes also do a separate chat only repo or server etc if cross-team agent chat access worth it and don't bother with private / public just disable agents in code repo
have also done some other variants etc. bit of pain to init config, does help with reducing the agent clutter, improve chat security boundaries in case any keys or sensitive env were in any chats if PII / sensitive data checker miss etc.
I cancelled my subscription over this [1].
1: https://x.com/m_herrmann/status/2092579839676948495
Indeed, the disclaiming of it being human-coded isn't the problem with their commit messages, them using it as a growth hack mechanism is. BTW to say it's attributed to Claude, I don't give that much importance, but I think it's useful to other coders to know what model was used. A simple assisted by AI would mostly handle the plaigiarism problem, because Claude's training data didn't come from Claude.
Interesting. Can you elaborate on why you feel strongly about this? It's easily removable--I don't think your characterization that it "forces" attribution is correct, it just defaults it.
It's deliberately hostile, in the way that Facebook is to its userbase. I guess Claude doesn't have nearly the grip on their users that Facebook does. Facebook can make it hard to communicate with (via WhatsApp) or stay connected with some of your friends or relatives, while Claude can just make you use the second-best model for cheaper (if Fable is the best).
who cares?
Some people who are in this thread might be interested in that view. A little more context would have helped it though I agree.
Beyond "branding", I don't mind this general direction.
There absolutely should be a fool proof undefeatable mechanism that identifies or watermarks code that was completely LLM generated.
If you use LLMs, own it.
"completely" LLM generated?
Hard boundaries there simply don't exist for most workflows, and I suspect if they did, it'd be at the expense of things ~nobody would want to give up.
you own the code, not the use of an LLM
The issue comments concerned about IP theft are deeply ironic.
Issue was closed with an robot-ai comment... Patetic...
This isn’t the only ridiculous thing Anthropic does to promote Claude. They still don’t support AGENTS.md
I really hate this attribution nonsense ...
On one hand I guess it is fine to automatically tag that the commit was AI-generated. On the other hand, this will happen even if you audit every line and add your own changes too, which means the attribution is misleading.
Having only 1 of the token sellers do this is not very helpful either. Its just free branding & advertising for Misanthropic.
> Misanthropic
How have I not heard this one before
It’s not meant for saying “this change was done by Claude” it’s meant to connect commits back to the session. Makes it way easier to jump back into a follow up change between sessions as the agent can reload parts of the context.
Anyone thinking this is about tagging AI generated changes and not just making the experience in Claude better is missing the point of the feature.
I get that. However, given that sessions are local and personal why tag it into the public source tree when it can keep a catalogue within the local session metadata with the exact branch, tags and refs. Sounds like easier less intrusive thing to do.
Generally just very useful to have it clickable or copyable from the commit is the thing?
Along with models being much better and more direct to follow something if it’s there. An agent will always read commits to look at a change, force it to look at commits and check another index just in case it exists is slower and pays more tokens, now you can just grab it.
But also mostly because who would actually care about this. It’s a sentence on the end of a git commit, in what way does this annoy people other than their prior feelings about AI tech.
Is this not an avenue to distributed distillation? Wonder how Ant squares that circle
Couldn't this be used by Anthropic to measure the usage of its product?
Anthropic is right here. You need to clearly define what commits are AI authored or you are conspiring to commit copyfraud.
https://en.wikipedia.org/wiki/Copyfraud
AI does not enjoy copyright protection. Pretending you created something written by AI is copyfraud.
> AI does not enjoy copyright protection. Pretending you created something written by AI is copyfraud.
Which court has made a ruling on this? The only decision I'm aware of was that some Gen-AI artwork couldn't be copyrightable if it was made without substantial input, direction, and scrutiny.
Once again, I will point you to the law,
https://www.congress.gov/crs-product/LSB10922
You did not write it, the AI did. Either the work cannot be copyright, or the AI owner (Anthropic) owns the code. Not you. In no case can you own code you didn't write. You guys keep arguing that it enjoys copyright protection without realizing this part. There is no "work-for-hire" default in the law for AI even if it did enjoy copyright. You don't own it. And I am pointing out neither should Anthropic.
That's an opinion from the US copyright office, not a law or a decision made in court. I think it remains to be seen to what extent using generative AI affects the practical application of copyright in larger projects: the office holds that the human-contributed parts still qualify for copyright, and I think there are a variety of possible interpretations and arguments that could effectively allow a largely AI-generated codebase to still be protected by copyright, even if technically big sections of it are not (in a similar manner as e.g. maps are protected, even if the underlying facts are not).
(I would also argue whether it matters in practice the other way: if something was AI generated, it is in principle cheap for someone else to generate it separately anyway: if you're wanting to copy AI generated code then there's probably some part of the ideas in it that are valuable beyond what the AI can output on its own)
(Also, this uncertainty does still mean that it's probably prudent to record which parts of your software are generated and which are not: it's a conceivable outcome that some company winds up in court and manages to argue that the human-generated parts are still covered under copyright, but are unable to point to which parts those are and therefore lose it entirely)
Yes, that's exactly what they were talking about. It says right in that page that sufficient changes can allow them to copyright the work, and it also says that their decision can be challenged.
And once again, you guys make this claim without reading the law.
>several provisions of the Copyright Act imply that it uses the word "author" only to refer to human beings, including provisions (1) vesting copyright ownership "initially in the author"; (2) limiting copyright duration to 70 years after "the author's death"; (3) providing for inheritance of certain rights by the author's "widow or widower" or "surviving children or grandchildren"; (4) requiring a signature to transfer copyright ownership; (5) extending protection to unpublished works regardless of the author's "nationality or domicile"; and (6) defining a "joint work" based on the authors' "intention" to merge their contributions in a certain way.
The law clearly, definitively, is talking about human authorship and nothing else. You are basing this concept that you can copyright AI written code on a law change that has not happened. It is very very unlikely that a law change can/will ever happen, because copyright law in particular is enshrined in international treaties like the 140 year old Berne Convention. This is a major reason why I was very much opposed to the TPP. Every new international copyright treaty is just more cement, making it completely unchangable. If you want to change this law, it will require cooperation from most of the countries in the world to do it.
To be clear, the Copyright Office is providing guidance and a particular (respected) interpretation of the law. Their opinions do influence courts, but they do not constitute law themselves.
Interesting. According to the definition in that Wikipedia link, "copyfraud" is claiming copyright on public domain content.
Does this mean a company can take code from an open source project that used AI, put it into their closed source code, potentially modify it, all without attribution?
Yes, that's exactly what it means. AI works are public domain by default. Copyright law requires the work is created with human creativity. The essential piece is to clearly define what parts are AI generated vs human written. This is an obligation placed on you by the courts, and if you don't do it, it becomes very very hard to argue that you wrote that code.
There is of course, transformative use, and the slew of software packages we used to depend on without acknowledging the original maintainers on every commit.
If anthropic believes in giving credit, why not give credit to all the content they stole?
Isn't that implied when it says coauthored by Claude? Would be waste of chars to name everyone else
Where can I see the list of everybody they stole IP from? Just saying "that guy stole it for me" isn't really giving credit to the real owners/creators.
I don't disagree with you on this. However, it is clear to me that a lot of copyfraud is happening in this industry right now. Just look at all the github repos with a) a licnese.txt and b) a claude.md in the repo.
Does Microsoft windows give credits to every developer who ever worked on it?
That’s just your opinion on your own characterisation.
Very much agree, and per https://www.jvt.me/posts/2026/02/25/llm-attribute/ there are a number of other benefits to attributing correctly - you know what parts of the code may be "tainted" by AI-generated code, and allow yourself to clarify to the reader what parts are AI and what are not
Copyright shouldn't exist and AI (especially open weight models) is accelerationism to this point
Everyone loves AI so much that they want to hide the fact that they're using AI.
Maybe using git notes for the conversation or conversation summary is better.
I don't think git-notes has the same history / change tracking. Use a .sessions dir or similar. I'm working towards adding them to my knowledge base so as not to pollute my git repo
I needed a reason to look for a different Harness. This could be it
Good way to find engineers who don’t know what they are doing and just fire them, from examining their sessions.
A lot of people say this is a reasonable feature, and I agree (I would strongly consider having it on), but that's not the defense they think it is.
- Having the feature available seems good.
- Having the feature default on is debatable, but not outside the realm of possibility.
- Having the feature default on, without any kind of release notes that say that it happens and link to the option to disable it is clownshoes "bring back the ghost of Steve Jobs to yell at the product team" type behavior.
Also codex cloud and others.
Luckily I've cancelled my subscription.
[flagged]
[flagged]
[dead]
[dead]
[flagged]
[dead]
Wouldn't think I'd be cheering Antrophic decision, but this is excellent.
They probably were thinking about how that serves them as an ad, not about transparency, but still.
I say they should go one step further and somehow make it harder/inconvenient to turn it off (big tech is good at this already, so...); or, maybe, hide that option behind extra feature under premium subscription or something.
P.S. The comments are pure gold. They don't like they can't fake authorship and have the audacity to be indignant!
As a Claude user this is just a benefit to me. Means I can grab the session and resume if I ever see a commit I want to jump back into, and Claude can read these sessions itself whenever I reference the commit, which helps a bunch.
It might come with advertising benefits for them but expect it’s mostly just because it helps their tools work and that benefits everyone.