More on Anthropic's Scion ripping off Mononoke | Hacker News Reader

More on Anthropic's Scion ripping off Mononoke

sunshowers.io

257 pointssunshowers91 comments

Redesigning source control at Anthropic - https://news.ycombinator.com/item?id=50073633 (996 comments)

Anthropic's Scion is heavily inspired by Mononoke - https://news.ycombinator.com/item?id=50094694 (149 comments)

Hi! I am the original author of the blog post and wanted to clear the air a bit here.

We credited Sapling because that's how Meta itself externally refers to the entire system, including the Mononoke server, and we wanted to follow that convention. We will add an additional footnote to clarify this. [edit: this is done]

We were inspired by many things while building Scion, such as the great ideas in Mononoke's Bonsai data model, and didn't intend to obfuscate that in any way. We were also inspired heavily by https://martin.kleppmann.com/2015/03/04/turning-the-database..., for example, which also describes the "derived data" concept using the same words.

We also did not just tell Claude to build us a version control system; we did a lot of research on what would work and not work for our use case and coming from an existing large Git repo (rather than Mercurial, as Mononoke did) required many novelties beyond some of these data structures.

The intent was to give credit to Mononoke without getting too deep in the weeds of comparing and contrasting each point of the data model where we diverged. There are many ex-Meta employees at Anthropic who were inspired by how joyful using source control was because of Mononoke, which deserves a lot of credit to you and the source control team there. We had that experience as a north star from the beginning while making decisions that map to our (different) needs.

It feels strange to put "-anthropic" in your Hacker News username. Never seen this before. What happens if you don't work there anymore?

They throw it away?

Just make a new account? Not like it costs money.

He might already have a main account but intentionally doesn't want to use it. I've noticed on X that OpenAI employees tend to post a lot more regarding their employer on their main X accounts than Anthropic employees do.

I don't use HN often, so I made an account for this purpose that would make it clear who I was in relation to the article.

You also register -openai -deepmind -meta and -xsi

Then there will be no more time spent trying to manage Orange Site public opinions regarding Anthropic.

Just a way to separate one's personal comments from their professional ones, like a work email address.

Why would someone respond to a post in a professional capacity with their personal HN account?

You ask that like it's not the absolute norm :P

Thanks for coming on to "take the heat." Do you have any internal tools for auditing the codebase/model outputs to see how much direct overlap there is with the inspiration?

As a human who is deeply familiar with the codebase, what percentage of Scion's codebase would you estimate is directly inspired by Mononoke?

Very little, overall.

This system was actually built incrementally over several months and did not start out as a full VCS -- it started as a read-only service that ingested data from our Git repo to take pressure off of CI and other tools needing to materialize checkouts for certain types of repo/graph questions. Even when we did decide to go all the way, we were operating in a hybrid Git/Scion world for months during the migration and the path to get there made the system and code very unique to our infrastructure.

This is detail that would have been interesting to read in the original post.

This is very similar to the way SCMQuery existed before Mononoke and eventually became served by it. The source control team at Meta was, and is, quite familiar with hybrid deployments of the kind you describe, though of course at longer timescales.

I'm finding this all really much ado about nothing. What does it matter how much overlap there is? I laud nickg for clearing the air because I personally would've just let everyone stew if it were up to me. All this scrutiny is happening because "big AI taking everything" and it's ultimately pointless. They made a tool that has a lot from an existing tool, yes. They also stated where some of the inspiration came from, which is directly related to the project the author is screaming about. Blog post author decides - against their better judgment IMO - to dip a toe in, and suddenly there's drag on that toe, for nothing. This is all just... wild.

Looking deeper, I think Ant really did copy the work. I don't think you should feel ashamed or defend it by saying that a lot of research is required.

Yes, copying takes a lot of research. It takes a lot of effort to collect OSS data, and a huge amount of research is needed.

As the longest serving and most senior member (now ex-member) of the Facebook source control team, I had no problem with your post. I'm just happy to see the ideas and technology helping move the industry forward.

Thanks for the response, Durham. I hope you're doing well. I've always appreciated your magnanimity, but I want to note that this belief is not unanimous among the current source control team at Meta.

There's a certain irony in the fact that Facebook did basically exactly the same thing when they made their post here: https://developers.secure.facebook.com/blog/post/2022/11/15/...

I see zero mention of srcfs or blaze.

Perfectly aligns with the V-Capitalist/Zuckerberg mindset: First you salvage them (like Facebook did salvage MySpace and others via interoperability), then you build walls (walled garden) to keep others from doing the same as you did, complaining about it via PR, and lobby law-makers to forbid interoperability to keep your money garden safe. Never mention who's shoulders you stand on or how you got there. Propel the myth It's all made by hand from the mud by the single 'genios' CEO-archtype - that sociopathic mindset infects the whole company and, if these companies are successful on the market, infects the whole economy, then society until we are where we are.

If the original idea was in those projects, Anthropic "forgot" to credit them too. One wrong doesn't give an indulgence for other's to copycat that wrong action later on.

To be clear, I don't believe either of these were wrongs. In my opinion, this is just a symptom of Facebook's PSC process training people to clout chase and demand recognition.

While your post sugar coats it in positive words: "joyful", "inspired", and so on, it just largely confirms the accusations in the blog post:

> However, after publishing the post I privately received confirmation that Scion intentionally and blatantly ripped off Mononoke, and in particular that there was overlap in personnel.

I come out of this that Anthropic actually ripped off Mononoke, but did so in a joyful and inspired manner. Maybe the prompt to Claude has a trailing smiling emoji.

Do you actually have a specific ethical line in mind that you think Anthropic was on the wrong side of here?

FWIW, I think they should've been much more open in their announcement and credited original teams instead of selling this as some kind of almost-autonomous Claude project.

"We took existing system we helped to implement at Meta and changed it a bit" doesn't sound as cool as "Redesigning source control at Anthropic", but seems to be closer to truth?

Yes, dishonesty. Flaunting this great technical achievement of their Mighty Labor Arbitrage Machine when it would have been impossible for them to have done without Mononoke already being open source.

Yes, they should have been clear that they ripped off an existing product instead of trying to make it sound like Claude came up with this amazing new thing.

I too can use LLMs to copy existing shit.

Thank you for responding. With the new footnote [1], I now believe the attribution is sufficient. I've added a link to it, as well as to your response, as an update to my blog post.

[1] https://www.anthropic.com/engineering/redesigning-source-con...

Milking this fake drama for all it's worth huh? Good to know you, the authority on the subject, believe it's sufficient now.

As if attributing Sapling didn't include its server, Mononoke. What's next, an author of some sub-module of Mononoke will be "outraged" he didn't get personally attributed?

This is all just your ego speaking. Good job taking advantage of the "anthropic bad" undercurrent, I guess?

> Update: Nick G. from Anthropic has responded on Hacker News, and sufficient attribution has been added to Anthropic’s blog post. I stand by my posts as of the time I wrote them.

This is such a weird turn of phrase. You _stand_ by your posts, implying that you currently stand by them, but then add "as of the time" that you wrote them, implying that you no longer stand by them? Either you stand by your posts or you don't; you can't have it both ways. It's just a little confusing.

It just means that Rain continues to stand by the posts each as a response to the circumstances at time of writing.

So you are claiming that if someone says “it’s raining” while there is precipitation, and then it stops, they can say it is “no longer raining” but aren’t allowed to believe it was raining before?

Seems to me that this is an inherent problem in all AI derived outputs. Take "inspiration" from everywhere else and give no attribution to the giants whose shoulders you build upon.

Maybe the next solution in search of a problem that anthropic and any other frontier data harvester should be looking to solve is accurate attribution to those "inspirations".

Do you give credit to every textbook you read and learned from in your entire life, every time you publish something?

Textbooks are not the source of the algorithms though, are they? They just usually republish them.

The Art of Computer Programming would like to have a word with you.

Hehe, This is how I know you're a programmer finding the edge case or exception.

If I have seen further, it is because I shat on the shoulders of giants.

(h/t amlib: https://news.ycombinator.com/item?id=49963529 )

Yet another case, when a company steals someone's IP (1), then that person/org has to complain which spending their valuable time for that, then they are getting blamed by readers for even raising the doubt, then they need to waste even more time for writing a laundry list of proofs that they were right the first time, and then maybe the corporation infringing on their rights acknowledges that "oh, sorry, here is the attribution" in best case.

It's like that joke, about praying to god for a bicycle, but then realizing that god doesn't work this way, so instead stealing a bicycle and praying for forgiveness. Seems to be an operating principle of all megacorpos these days.

(1) Not complying with a license to commercially using someone's IP is stealing said IP; license requires attribution which wasn't given so that was the infringement.

> my default assumption now is that Anthropic’s public engineering presentations are largely bad-faith ripoffs. That’s a shame.

Couldn't agree more.

I got nervous when the company claiming to be working on the most impactful technology in human history takes time away to instead brag about writing a slightly more efficient version control system.

The fact that they didn't actually build the side project themselves, and ripped it off from someone else also calls into question their good faith mission to shepherd humanity into a new era...

Is it surprising? The bulk of their work is largely derived from existing works themselves, nothing original, everything probabilistically paraphrased and then post-trained by lowly-paid, ESL workers resulting in the really annoying, mic-droppy writing output. It will never be original, it will never be new, it will always only be derivative.

They have not made anything revolutionary, they have only ever made evolutionary derivatives of medium-to-high quality with no accreditation or financial support for the original authors (at least not without being forced into a class-action over it).

> They have not made anything revolutionary

Hard disagree. Claude and other frontier models are clearly a huge accomplishment, are already changing the world, and continue to get better.

Despite the real innovation, the way Anthropic communicates their work often comes across as disingenuous.

"Our models wrote a new VCS" is very different from "Our models adapted existing VCS to our bespoke needs". Both are impressive, so why over inflate the claim?

Same thing happened with the Bun Rust rewrite.

LLMs existed long before Anthropic came to the scene. Tensor logic was not created by them, only implemented. An engineer does a lot of cool good things, sure - but it's the theory grounding it that's revolutionary and that was not private enterprise driven.

The theory did not do anything revolutionary. The products did.

This is very much a chicken and egg problem except we know which came first.

Yes it was! Transformers came out of a Google lab. In the 2020 Google I/O Sundar Pichai even demoed a transformer based chatbot smart enough to larp as a paper airplane.

These things are the work of many hands, mostly privately funded, and it's indisputable that Anthropic has advanced the field.

Open source Mononoke and EdenFS don’t really work outside of Meta. I looked into it some years ago for a large monorepo, determined it wouldn’t work, and shelved it. If I had a team of people and/or agents available we likely would have built it ground up instead of adopting Sapling’s codebase.

Seems a bit silly to raise all of this fuss over an MIT licensed project that is both open source and not actually usable. What do you expect people to do with the source code, admire it?

In some cases, yes! What would you put in a museum of programming?

> Even the name “Scion” reverses the syllables of “Bonsai” and drops the “B”! This might be a coincidence, or might not.

huh

“The meanings overlap too. In horticulture a scion is the cutting grafted onto a rootstock, and grafting is a standard bonsai technique, used for example to add branches to pines and junipers. So two gardening words line up in sound and in subject.”- Claude

Lol

Reminds me of Glenn Beck’s tirades on TV in front of a blackboard where he’d stitch together new tinfoil conspiracies in real time.

Yea drop the B, replace a with c, and move all the letters around… other than that it’s the exact same.

Syllables, not letters. Bonsai -> sai-on -> Scion

Of course, bison was already taken

Recent and related:

Redesigning source control at Anthropic - https://news.ycombinator.com/item?id=50073633 (996 comments)

Anthropic's Scion is heavily inspired by Mononoke - https://news.ycombinator.com/item?id=50094694 (149 comments)

Re-designing a modern source control system for open source projects, you need to consider millions of developers have coding agents running on their computer and/or have Claude Code/Google/OpenAI coding agent accounts. And evidence from the Linux repo and many other GitHub repos, the open source projects are overflowing from coding agent submitted bug reports. If I was accepting a bug report and/or patch, I would want to know the stated problem and reasons why it's a problem, the desired state or fix and reasons why it's preferred, and if a patch is supplied, I would want verifiable proof of provenance: the signed session from the coding agent, including the prompts used to generate it paired with the changed code. This way open source maintenance can become truly decentralized.

That totally sucks. Thanks for providing more details.

Makes you wonder about the mathematicians who are also saying similar things.

Anthropic's Scion is heavily inspired by Mononoke

https://news.ycombinator.com/item?id=50094694

Picasso apocryphally said "good artists copy, great artists steal"

> Imagine how they must feel about their work being deliberately copied wholesale and intentionally presented as someone else’s.

Nobody copies from a mid-tier product.

As someone who worked in open source till recently, I've seen a copycat as validation that my idea was better than what they could come up with on their own.

All those people now have evidence that their approach is the best for an AI native future.

The interesting point is that there is overlap in personnel, so this is less of copied homework, more that Anthropic hired some good people who knew a good solution that worked.

But from a pure AI lens, this is what we're doing - putting great ideas into one system and now everyone can drink from that well.

The bad part is that the well has a coin meter next to it, rather than it being a commons. So the fruits of that is not distributed to the people who planted those seeds. This is also happening too fast that the people whose work is getting redistributed to others are still not done being compensated for their work.

"Good artists copy, great artists steal" doesn't seem to me to about how wholesale the copying is but that the copy exceeds the original in creativity, beauty, or some other quality so much that it displaces the original, and makes the original seem like the inferior copy.

In that case copying something that is merely mediocre in most respects but has one outstanding quality, and generally exceeding it, makes sense.

I'm not sure any of that applies here except maybe that Anthropic might be more poised to make anything they publish suddenly very famous.

> “Our LLM built this thing, productionized it, and dealt with all the attendant challenges over 5 months” is a significantly more compelling narrative than “the source control team at Meta had spent 10 years building and productionizing this system, and we were able to point our LLM at it and rebuild it in 5 months”.

This is, unfortunately, the largest underpinning issue with a lot of software currently being released from AI-heavy engineering workflows. Even if it wasn't done explicitly, (almost) everything modern LLMs spit out is built on top of decades of prior failures by humans; even if something is "novel", it's trivially easy for the LLM to start with excluding all the paths that failed before (that humans were generous enough to share freely)

tldr; almost everything these days is a copy, even if we don't know what it's a copy of (yet).

Crediting Sapling for 'hard-won lessons' is like crediting the answer key for helping with your homework.

> Building a write-ahead log-style Merkle DAG was very far away from what anyone else was doing in 2018

That completely ignores the fact that Google's Piper had something very similar working already in 2012.

Thanks for the comment. To the best of my knowledge, Piper is not Merkle DAG based, and instead uses a (completely linear?) sequence number similar to Subversion. I believe they ended up faking those hashes for Fig. I'm not actually sure what they do for Jujutsu compatibility, maybe they've added a Merkle layer since.

We were compelled to model things as a DAG to be able to handle the occasional merge commit (this makes the algorithms a lot harder), and we wanted the Merkleness for its desirable properties such as being easily cacheable and straightforward to do integrity checks on. We wrote a bidirectional conversion layer between the Mercurial and Bonsai data models, with great attention paid to complete reversibility. (That's how I got into property-based testing, by the way! My coworker also used Alloy to formally verify the data model.)

The closest equivalent is probably Monotone, actually, though it only had one tree/manifest format which was baked into the changeset, while Mononoke got rid of that and _only_ relied on the changes, flipped the pointer directions, and generalized that to arbitrary kinds of derived data. (This was at the cost of having to maintain mapping layers in a database.)

Just because you might find this interesting: Scion deviates significantly in this regard in order to deal with merges; coming from git, we land code to our main branch via merge so it's a common occurrence.

We do discourage (but not block) some trickier DAG things you could have done in git and are probably moving towards squashing and rebasing for landing since many things become simpler without merges, but we simply had to deal with merges (and handle git compatibility while migrating) to pull this off. Particularly, full DAG merge bases add a lot of complexity and it gave us some concrete ideas on how to improve git's commit graph data structures by an order of magnitude or more.

I'm not convinced that Merkle DAGs are useful or worth the bother. Piper's system of linear indices has a few nice properties:

1) when a defect is identified in a library, the extent of the defect becomes an interval [CL that introduced it, CL that removed it] and it makes it trivial to test whether a compiled artifact is affected by that defect; more generally the UX is a lot better

and 2) it makes whole-repository rewrites trivial, so that existing commit sequence IDs are preserved while retroactively removing sensitive files that shouldn't have been added to the repo. Git and Mercurial make such rewrite a huge pain in the arse, while this is an essential feature to have in an enterprise VCS.

These were both solved as part of the data model via globalrevs and obliteration respectively. But that is besides the point here? The next time you're designing a source control system you're welcome to pick your favorite points in the design space.

> These were both solved as part of the data model via globalrevs and obliteration respectively

Is there some paper or other technical publication regarding these concepts ?

I'm not aware of anything published, but the code is open source, and if you're so inclined you can probably point a clanker at it. The second one is called "redaction" in Mononoke, I believe (obliteration is a different term for it, as is censorship).

You are correct that integer sequence numbers and redaction are both reasonable things to want -- we just believed (for a variety of reasons) that the core should be Merkle and that these things should be built on top.

> if you're so inclined you can probably point a clanker at it

I'm not. Too bad there's nothing meant for humans.

How do you feel about the adobe cloned products?

The entire thing is beyond confusing. So Anthropic refers to an open-sourced project and credit it in the original article.

I don't know how the author is pissed off. Only thing the author aware is Claude agents researched Mononoke solution, then it claims a closed source repo is a plagiarism case.

I don't get it, and I'm seeing more and more cope here.

My read of the author's position is that the people who built Mononoke at Meta did all the hard engineering work that made Scion possible, while the work that happened at Anthropic was mere code-monkeying unworthy of note. This is decidedly not the sense you would get from reading Anthropic's blog post, so they feel that that post misrepresented what happened. I am skeptical that this is really an accurate characterization, but I can see how a person might get there, especially if they were involved in the work at Meta and so have a natural bias towards weighting its significance more heavily.

Every ducking time I hear Anthropic or OpenAI did some great engineering it either turns out to be shit (c compiler) or stolen (this). Bravo! LLMs are such a marvelous technology, but being developed by such dushbags that I suddenly root for Meta. Disgusting.

Honestly, there have always been parallel inventions that came out from different parts of the world at the same time, without knowing of each other.

From an information-systems perspective, many systems have just optimal solutions that all developers will eventually converge to if they optimize on the same problem.

It's well known that AI companies shamelessly steal whatever to drum up hype for their company. From their perspective you the problem for trying to stop their hype train.

An employee at one of the worst companies in human history (META)--whose modus operandi is to destroy the social commons--complains that another of the modern era's potentially worst companies (Anthropic is, after all, offering a 10% chance of human annihilation) stole its data/code/something without proper attribution.

Rhymes with when Anthropic and OpenAI were mad that folks were distilling their models after Anthropic and OpenAI violated every intellectual property law they possibly could to build said models.

What you're watching is basically if 3 Warlmart variants showed up in your town to fight over who gets to run the mom and pop shops off the internet.

Yeah, I simply could not give a shit less about any of this. Everyone involved has furthered the agenda of awful companies that harm society.

The author hasn't worked at Meta for eight years

Then I guess he never worked at Meta! https://about.fb.com/news/2021/10/facebook-company-is-now-me...

Also, sorry, but Facebook has been evil for way longer than eight years.

I mean they were evil from the beginning; my point is that she has has had plenty of time working at a non-evil company to not be dragged down for it. One could serve less time in prison for armed bank robbery

tbh i think anyone who has worked at meta probably should serve that much time in prison...

Working at meta is essentially identical to serving time in prison, so.

(I left Meta in September 2022, so I haven't worked there in four years. My current coworkers can attest to the fact that I'm not proud of working at Meta, but I am proud of the world-class nature of what our teams built. In any case, I don't believe my personal feelings about Meta have any bearing on the matter.)

Although I'm totally on board of the high-level points,

> This remains yet another shameful example of large AI companies appropriating the commons and presenting others’ work as their own.

If you work at Meta which is, you know, one of large (aspiring) AI companies, a bit of self awareness might help boost your credibility a bit more.

How does Anthropic justify blatantly ripping off another company's work like this, especially without credit, then keep on whining about "distillation" when they suspect anyone else is merely using the output of their LLM in ways they don't like.

The other interesting part of this is the amount of time and effort it took Anthropic to do this, despite the fact that they were just copying an existing design - 5 months and (from another thread) 7 humans... Why couldn't one of their coding models/agents have done this by itself given the detailed specification? It's understandable that an LLM can't design a system like this by itself, but a bit surprising that even given the design is apparently can't code it up by itself either (else why did it take 5 months!).

Honestly I think this is a really bad look and sour grapes from the OP. Sapling is GPL. You open sourced it. What does “rip off” mean here?

If you think the license was breached and code was copied, that’s one thing.

But claiming ownership over an abstract idea in a project you open sourced seems epistemically confused to me. You don’t get to do that, there is no obligation from anyone to credit what OSS projects they drew inspiration from, as long as they comply with the license.

Having said that it’s nice to give credit to your inspirations, and Anthropic very clearly did:

> Source control for huge monorepos has prior art, notably Google’s Piper and Meta’s Sapling, so we were able to build Scion with their hard-won lessons in mind

This is just needless drama, nobody at Anthropic did anything wrong here.

> Sapling is GPL. You open sourced it. What does “rip off” mean here?

Exactly. This is the same type of nonsense nerd drama that happens sometimes between various Linux distro fans.

There is a distinction between legal obligation and following basic norms. Save for extreme circumstances, there are not laws against being a jerk or being careless. Still people reasonably call on others not to exhibit jerk behavior or carelessness.

"With hard-won lessons of [an encompassing system]" does not give the impression that the underlying data model for changes was directly adopted. Neither does your language of "abstract idea" and "inspiration" adequately reflect the extent of borrowing.

Bonsai appears to have been, well... grafted; and it wouldn't surprise me if "Scion" was suggested by Claude specifically owing to its connotation of grafting. In any event, I imagine Claude was quite pleased with the name. To reiterate grafting in itself is no problem, but it warrants proportionate attribution, certainly so in public communication when relevant to what has been accomplished.

Attribution is a reflection of one's values. Dismissing the requests of others for attribution as "needless drama" is also a reflection of one's values, independent of agreement on facts.

The Anthropic article [1] has author names as initials or first names,and I thought it was to prevent talent pouching, but now other possible intentions come to my mind...

https://www.anthropic.com/engineering/redesigning-source-con...

Reply on news.ycombinator.com