Open source chess developers warn about a commercial engine based on Stockfish | Hacker News Reader

Open source chess developers warn about a commercial engine based on Stockfish

lichess.org

285 points1337shadow130 comments

I'm glad the group spoke up. Free software or no, details matter, and in this case the details point to false claims and unethical behavior:

> It is sad to see claims of innovation where there has been none, and claims of improvement in an engine that is weaker than its open-source origins. It is also sad to see people appropriating the open-source work and effort of others and claiming it as their own.

> Everyone is permitted and encouraged to modify and improve code from Stockfish/Leela while giving credit; that is the intent of open-source software. Everyone is allowed to copy Stockfish/Leela and sell them, provided the terms of the Stockfish/Leela license are met. But don’t pretend that the product being sold is something it isn’t.

ChessBase is selling this Stockfish clone for €99.90. They have also disabled comments on all of their articles so far about Fat Fritz 2. Not a good look.

Related Article on Lichess (which is also open source) about Fat Fritz 2: https://lichess.org/blog/YCvy7xMAACIA8007/fat-fritz-2-is-a-r...

They are a sketchy company, from the over-zealousness to defend their "IP" to the price they demand for their yearly software, to the creepy vibe their articles about young female players have. So no sympathy from me, go stockfish.

Im looking at the diff, they doubled the size of NN and added themselves as additional authors in addition to the original authors and released source when requested. I.e. they gave credit to original devs, I don't understand where they violated anything. NN weights is data not code.

Yes but read the article. It's the way the "author" presented the work as novel that's indefensibly hucksterish

Do you have any links for these things?

People are never just ONE type of awful. It's almost always a package deal.

It's also just an appalling piece of software. I wish the market for competitive chess were bigger so someone was incentivised to dethrone them.

Thanks for the link. That git diff is absurd: 9 changed files with 21 additions and 30 deletions[1]. It really is just Stockfish using different weights and changing the name... The fact this Albert Silver fella did the same with Leela Chess Zero two years ago is even worse, no remorse!

[1] https://github.com/official-stockfish/Stockfish/compare/550f...

This comment was originally posted to https://news.ycombinator.com/item?id=26188485, which points to https://blog.stockfishchess.org/post/643239805544792064/stat..., a related article. We merged the threads.

Feel free to express your disgust by leaving feedback for the ChessBase editors at e.g.

https://en.chessbase.com/feedback/fat-fritz-2-best-of-both-w...

ITT: People missing the point by making it about software licenses, when the real issue is lying about the product.

ChessBase could easily comply with any open source license by linking the full source code, including the method they used to generate NN weights (if they think the license applies here as well), in a footnote.

It would still be a scam, because their claims of superiority, originality ("almost from scratch"), and innovation (see the funny part about technology from Japan) are wrong, and purposely so.

What are these _open source_ licenses, which require you to link to the full source code? I might not have seen such a license yet. Usually "open source license" equates to MIT license and BSD licenses, which do not require you to link to the source code.

AGPL?

Not all 'open source' licenses are 'non-copyleft' or 'permissive' licenses.

False claims of authorship (essentially plagiarism) are not entirely outside of the domain of copyright.

https://en.wikipedia.org/wiki/Attribution_(copyright)

For instance, let's look at the BSD license. It requires copyright notices in source code to be preserved. That means that it reserves the attribution right.

The original license required them in executable code; the two-clause version relaxes it; it doesn't demand any expression of attribution in the compiled code.

However, it does not appear to fully waive the attribution right; it requires the copyright notices and license clauses to remain in the source code. The compiled code is still the original author's copyright; nothing in the license says that it isn't.

If someone claims they developed it, it's infringement on the attribution copyright, even though they have the license-granted right to redistribute the code without making such deceptive claims.

IANAL, but if you're going to plagiarize, I suspect the safest bet is to pick something that has lapsed into the public domain, not something with copyrights and licenses in it.

Many commenters are missing that the issue is not that an Open Source project was repackaged as a commercial project, but that the open source origin has been hidden.

> but that the open source origin has been hidden.

What is the specific licensing issue here?

If this is about them not liking the product, fine, but don't dirty the open source ethos unless they are breaking the license. Part of the problem is the current HN headline perhaps.

It's GPL: its not attributed and the binary is available for download while the code isn't open sourced.

Code is open source (given when requested as per GPL [1]) but not the net weights.

I don't think under GPL the net weights have to be open sourced.

[1] https://blog.stockfishchess.org/post/643239805544792064/stat...

There seems to be confusion on what the issue is.

IANAL, but I don't think it's clear that the net weight can be keep secret. From https://www.gnu.org/licenses/gpl-3.0.html

> “Installation Information” for a User Product means any methods, procedures, authorization keys, or other information required to install and execute modified versions of a covered work in that User Product from a modified version of its Corresponding Source. The information must suffice to ensure that the continued functioning of the modified object code is in no case prevented or interfered with solely because modification has been made.

The specific licensing issue is that both are under GPL v3.0. Which means that you can sell a derivative, but you have to clearly advertise that it wasn't your work, provide source upon request, and so on.

They did none of this and so are in copyright violation. And so should be sued.

Is this the Ben Tilly that paints and lived in Seattle moved to ???

Hello Ben!

Sorry, different Ben. I grew up in Victoria, BC, never lived in Seattle, and have only painted a handful of times in my life.

Hmmm, ok. Funny, I grew up in Port Angeles.

...and if they're breaking the license, don't "name and shame"; send them a C&D. And if that's ignored, sue them.

Seriously, there's some attitude in open source circles that looks down on using the law to achieve compliance. Why bother putting up such a carefully thought-out legal document, and then whining on a blog when it's broken, instead of using it for its intended purpose -- a court of law?

Quicker, cheaper, and probably feels good in the moment. The law can work if you have time and money.

Also, due to the ill view of patents on this forum, I imagine people here are simply put off from using the law and thus opt for the dopamine hit.

Public shaming can be remarkably effective and does not preclude legal action later on if shame itself is insufficient.

Free software enthusiasts ought to know better than to conflate patents and copyright.

> Also, due to the ill view of patents on this forum, I imagine people here are simply put off from using the law

I'm not talking about HackerNews in particular here. There's a deep divide in the Linux kernel community, for instance, about GPL enforcement. Some people, often represented by the Software Freedom Conservancy, believe in "speak softly but carry a big stick": i.e., try by negotiation to get violators into compliance, but sue when negotiation fails. Other people (including, IIRC people like Greg KH) have at some point openly stated that they would never consider a lawsuit an appropriate course of action.

I'm not sure where the Leela Zero and Stockfish people stand on this; I just sort of assumed from the tone of the article that they were in the "never sue" camp. Maybe they're in the "I didn't realize we could do that" camp, in which case I'd strongly recommend they get in touch with SFC.

> ... send them a C&D. And if that's ignored, sue them. Why bother putting up such a carefully thought-out legal document ...&c&c&c... instead of using it for its intended purpose -- a court of law?

Because

* legal action is expensive, time-consuming, and may be simply impossible (or equivalently impracticable) depending on where you, or the other parties, reside/do business; and

* legal documents do not in actuality exist to enable someone's courtroom drama fantasies. In anticipation of civil disagreement, competent lawyers will advise that adversarial proceedings are a last resort, not a first measure, and draft documents accordingly.

> Many commenters are missing that the issue is not that an Open Source project was repackaged as a commercial project, but that the open source origin has been hidden.

That is not what Joost VandeVondele, current maintainer of the Stockfish project, posts:

> Recently, ChessBase has started distribution and sales of the Fat Fritz 2 chess engine. This chess engine is a Stockfish derivative, with a few lines of code modification (engine name, authors list and a few parameters), and a new set of NNUE net weights considered proprietary

> Selling Stockfish derivatives is possible with the GPLv3 license we grant, but not without requirements. In particular, the license states that if one redistributes a program derived from our work, the corresponding modifications of our sources and all information needed to build that program must be made available. Only after explicitly informing Albert Silver (the author of the net in Fat Fritz 2) of a license violation have matching C++ sources, but not the net weights, been made available. Obviously, we condemn the approach taken.

https://blog.stockfishchess.org/post/643239805544792064/stat...

How is this not exactly what the parent comment says?

This article is light on the details about this being a GPL license violation or not. The final paragraph also makes it sound like Stockfish is under a permissive license (only requiring attribution) when in fact it's licensed under the copyleft GPLv3.

It was, but according to Wikipedia they sorted that out by moving the proprietary neural network out of the executable. If this is correct, it means that the source is available and the nature of the changes (including removing the original authors) should be visible.

https://github.com/official-stockfish/Stockfish/blob/master/...

Stockfish _is_ currently licensed under GPLv3. Chessbase is allowed to sell a modified copy, but only under the same license and they have to provide the source.

The neural network is just data, but if it's part of the executable you need to provide the raw data that is then linked into the executable.

I think it can be argued that the set of weights of a neural network is a program, but not “the preferred form of the work for making modifications to it”.

That would mean the training set and training algorithm (and random number generator seed) would have to be distributed under the GPL, too.

I don't think it is always feasible to make the learning process entirely deterministic, especially for distributed reinforcement learning.

Binaries weren't historically deterministic either. This isn't a problem.

They did the same thing though. Two differently trained neural networks will have different outcomes, in this case they'd pick different chess moves in some special cases.

I’m fairly sure there were (maybe still are) linkers that, given duplicate symbol names, would happily grab one of them without checking whether they had the same code.

Add in parallel linking, and program behavior can change.

The stockfish blogpost is quite a lot clearer: https://blog.stockfishchess.org/post/643239805544792064/stat...

ChessBase just posted a new article by A. Silver at

https://en.chessbase.com/post/how-a-neural-network-is-made

where the author tries to argue that doubling the neural network size in StockFish makes for an appreciable strength increase. However, only blitz results are shown. Looking at http://computerchess.org.uk/ccrl/4040/ we see Fat Fritz being only 2 ELO stronger than SF at blitz while losing out on longer time controls. This questionable tradeoff makes for a very poor excuse to charge $100 versus the free SF.

Public comments are once again blocked.

Some stories from ChessBase Chess News describing these products, roughly reverse chronological order:

https://en.chessbase.com/post/chess-and-artificial-intellige... ChessBase founder Frederic Friedel talks about Fat Fritz

https://en.chessbase.com/post/interview-with-albert-silver-h... 1 hr video interview with Silver about his "path to inventing the [Fat Fritz] 2.0 engine"

https://en.chessbase.com/post/fat-fritz-2-0-the-new-number-1 "Fat Fritz has revolutionised the world of chess engines."

https://en.chessbase.com/post/fat-fritz-2-best-of-both-world...

https://en.chessbase.com/post/fat-fritz-ingenious-without-st...

https://en.chessbase.com/post/fat-fritz-and-the-1200 has a section "What differentiates Fat Fritz from Leela?"

https://en.chessbase.com/post/fat-fritz-defeats-stockfish-ma...

https://en.chessbase.com/post/fat-fritz-what-on-earth-is-tha... 2019 - Albert Silver describes in detail how he made his "AlphaZero clone" Fat Fritz

Have there been any court cases about whether neural net weights can be included under software licenses?

Copyright doesn't make a distinction between software and other text, but GPL3 probably does, and maybe allows changed weights without them being released (unlike software, where it's explicitly stated). I didn't read the license though.

No, but copyright (in the US and US-influenced countries) does make a crucial distinction between work that is an expression of human creativity and work that doesn't, even more strongly since Feist. If the weights are a dump from some kind of ML training algorithm, it's likely that they're not copyrightable under US law.

This (to its logical absurd extreme) would mean that compiled binaries are not copyrightable and only source code is.

The GPL is clear here: you must release sources in the preferred form of the work for making changes in it. For a neural network, that is the training data, models and scripts - unless you really are just editing the final weights.

For better or worse, US courts have not interpreted the Feist precedent to mean that compiled binaries are not copyrightable. I'm not sure if any defendant has tried that argument when admitting to copying binaries; it doesn't seem likely to be a winning argument.

Are neural net weights code or configuration settings? Are email connectivity parameters code or configuration settings?

How would a judge distinguish the two?

The analogy of "email connectivity parameters" is way too simplistic. Neural network weights are a lot more like VHDL code on FPGAs that enable/disable individual gates.

The NN engine is just a substrate. It can't do anything on its own, and what it can do is vastly altered by the specific database of weights that are applied.

Additionally, as a proportion of what makes a NN system work, the weights are the vast, physical bulk of it.

I don't think any legal team employing a real ML expert witness would have any trouble arguing that the weights are the code.

It sounds more like the weights are data like an image, and the engine is the image viewer.

Or like the bitstream to an fpga not the vhdl.

vhdl really is source, so it comes down to if nn weights really are more like vhdl or more like the bitstream.

It seems like the closest analog to a source for nn weights is training data and scripts.

In that case, it seems fine to license the weights seperately, just like the source to an image viewer or audio player, yet not some image that is displayed in it.

I'm not so sure about shipping them together under a single name as a single product. Maybe that act causes them to be obligated to provide the source to generate the nn weights?

I would love that, not because I value their data but because it would mean it was only their own attempt to mis-represent that now makes them have to give away their store. If they had simply packaged up their fork of Stockfish plus proprietary nn model honestly and in good faith, there would be no problem. Maybe they couldn't get $99 for that from very many people but, oh well!

I would have assumed that most people interested in chess engines are a pretty savvy and will do their research before buying anything, making the whole "create a sub-par ripoff" a rather ineffective strategy.

Not really. Plenty of middling to world-class chess players aren't technically literate, but use chess engines for preparation. And Chessbase is the biggest name in chess software, though they've always had a slightly scummy approach to marketing.

As long as the license allowed it, I'm not seeing the issue. Sure; it's a less-then-advertised product, but the world is full of those.

If people want to drop 100$ on something that doesn't work as well as the free alternative, more power to them. People do it everyday in the real world.

The issue is it is fraud

While this is something I would expect from a small random company, ChessBase is pretty big in the chess world.

For them to promote this "new" engine on their shop[0] feels wrong, I would expect them to meet certain quality threshold.

Lichess does mention this in the bottom of the post, which has the same feeling:

[0] https://shop.chessbase.com/en/

"It is sad to see claims of innovation where there has been none, and claims of improvement in an engine that is weaker than its open-source origins. It is also sad to see people appropriating the open-source work and effort of others and claiming it as their own."

The license allows to use the code, it does NOT allow taking the credit for things you did not do.

There's way too little credit given in open source in general, and I'm glad to see people fight back again (it was more customary to fight back before 2010).

And I might be the only idiot here, but I tried to run stockfish from source and it was not an instant process. In addition, there isn't really a front end, you essentially end up with a CLI to input moves.

And for those who say, hey you, there are prebuilt binaries, just download them! Check out this message [1]

" The binaries at the top of the table are fastest, but may not support all CPUs. If you don't know which CPU you have, you can go down the list and pick the first binary that does not crash. "

Yeah okay, a pragmatic set of instructions that can't fail...

From this I think we could see why a product built on top of stockfish might make sense, even if the linked product might not do as such.

[1]https://stockfishchess.org/download/

Because the engine is not an interface. You can have then engine you want on any interface that supports it. I suppose the easiest way to play stockfish is to click "Play with the computer" on lichess.org, you don't even have to signup!

Then, you will find a /lot/ of GUI interfaces for your favorite OS (Linux, Android, BSD, Win/Mac...), and probably it will have Stockfish included.

As such, you don't have to download or build stockfish, basically any free chess app that you will find will make it available to you. I can think of no reason to make a free chess interface without Stockfish.

There are literally hundreds of FREE stockfish packages with frontend all good to go. To ask 99$ and lie through your teeth when it comes to authorship and performance is a ruse to rip people off, with the help of a famous name, plain and simple.

> This chess engine is a Stockfish derivative, with a few lines of code modification (engine name, authors list and a few parameters), and a new set of NNUE net weights considered proprietary.

https://blog.stockfishchess.org/post/643239805544792064/stat...

You completely missed the point. If people buying this KNEW it was just a free engine with minor tinkering (that apparently even makes it weaker), they would most definitely reconsider. It is the false advertising (https://shop.chessbase.com/en/products/fat_fritz_2, https://en.chessbase.com/post/interview-with-albert-silver-h...) and pretending that is morally bankrupt.

Just like dropshipping, it may be legal but is amoral to many people. And informing the general public is important.

There is no way this is the outcome, but I’d love to live in the world where ChessBase get a proper commercial license to Stockfish to sell to people who don’t want to click three buttons to set it up themselves, in exchange for permanently opening up their database to the public.

Can somebody explain to me how NN is used in Stockfish? I always assumed that some heuristic rules are used by Stockfish to evaluate the position, and I think this was actually the case not that long ago. So, does this mean heuristics are now being completely replaced by NN evaluations in Stockfish, all the prior work being completely thrown away?

In fact, isn't the position evaluation pretty much the only thing that makes Stockfish different from Leela, the rest being just some tweaks on top of traversing the decision tree? And if it now seems that NN is better at this after all, what is the purpose of having these 2 engines at all, shouldn't all the future work just be moving over to Leela?

There's an explanation here: https://www.chessprogramming.org/Stockfish_NNUE

It uses clever integer arithmetic in the neural network, so you don't need a monster GPU to get good performance. I believe as of now it's using both the nn and the classic heuristic-based system for different situations.

It doesn't answer any of my questions, though. In the end, it doesn't matter how exactly that NN works, in terms of architecture, weights, framework, etc. Given there is a common API, one position evaluation module can be used as a drop-in-replacement for some other implementation of position evaluation module. I mean, it kinda made sense that Leela was a different project, because it wasn't clear back then if a community will be able to train an NN which can surpass heuristics-based engine, and heuristics was at the heart of Stockfish. But if it turns out to be that heuristics-based approach has lost to NN, it isn't that clear anymore why these have to be 2 separate projects. At least, in terms of the rest of the engine.

> I believe as of now it's using both the nn and the classic heuristic-based system for different situations.

That's what I'm actually curious about. How does it combine these 2, precisely?

Edit: actually, reading more carefully the "Hybrid" section it kind of answers the last question. But still makes me wonder if the rest of the 2 engines cannot be combined.

Stockfish has two key components:

(1) an evaluation function, that quantifies black/white advantage in a chess position.

(2) a search algorithm, that iteratively checks move sequences from a starting position using (1) and returns a recommendation of the best move to play.

Historically (1) was handcrafted, but it recently switched to a 'NNUE' neural network architecture (now it using different evaluation functions depending on how balanced the position is).

This is a good video if you want to understand how Stockfish works.

https://www.youtube.com/watch?v=pUyURF1Tqvg

"But if it turns out to be that heuristics-based approach has lost to NN, it isn't that clear anymore why these have to be 2 separate projects. At least, in terms of the rest of the engine."

Leela's evaluation function is a deep neural network, inference can only be carried out efficiently using GPU, whilst the Stockfish system has been optimised for a CPU-based computer architecture. The innovation of NNUE is that inference can be performed efficiently on a CPU, partly because the network is much smaller, and partly because the architecture is designed to be efficient at recomputing evaluations of only a single piece has changed.

Clicking the link with lichess installed just opens the app. Long-press to open in a new tab.

FYI the lichess mobile apps have been abandoned in favour of the web view, if I recall correctly.

Website is perhaps recommended, but the app still works great and is updated from time to time.

Oh that’s good to know. I just be running on outdated information.

I was hoping for this (the app doesn't work on my old iPad, and I wanted Over-the-board play, but its an app-only feature).

Hope that comes to the website soon.

The iOS app got an update just a week ago.

I didn’t realise! My info is obviously outdated. Good to know.

Copyright issues aside, you would think claims of innovation would be taken with a grain of salt until this engine competes with, and beats Stockfish and Leela at the TCEC (top computer engine championship).

Please read till the end:

Everyone is permitted and encouraged to modify and improve code from Stockfish/Leela while giving credit; that is the intent of open-source software. Everyone is allowed to copy Stockfish/Leela and sell them, provided the terms of the Stockfish/Leela license are met. But don’t pretend that the product being sold is something it isn’t.

so sad. I will not use chessbase for anything lichess it is

Stockfish is under the GPL (IIRC) so I wouldn't be that worried.

GPL means you can put a version with proprietary modifications in the cloud and sell api access, without contributing anything back.

Right. I thought it was an app so my bad.

It is available both for download and in cloud. So they would have to pull the download if they dont want to release source.

Pulling the download isn't sufficient if it has already been downloaded - either they have to offer the source to everyone who downloaded, or they have infringed copyright by this distribution and can be justifiably sued for that.

However, this article implies that nobody really cares about them releasing the source, as it seems that it does not have any interesting improvements.

Yeah if they pull it they need to release current modifications, but not future ones.

And they only need to provide the modifications to people who downloaded it I think.

Yes, but they need to provide the code under the same GPL license, so in practice it would be available to everyone in the world as long as a single person who downloaded it wants to share it futher.

Would it finally be the case where a GPL violator is taken to court?

Hopefully. I also hope this gets spread throughout the online chess community so that ChessBase gets their name dragged through the mud, as they rightfully should, for this blatant plagiarism.

Taking the work of others, erasing their name and inserting your own, is the kind of crap we expect from lousy cheaters in high school, not businesses who want to establish themselves in a market based on a centuries-old global institution.

> Taking the work of others, erasing their name and inserting your own, is the kind of crap we expect from lousy cheaters in high school,

Or Amazon Web Services

The market and the history of the game is precisely the reason they became like this. It's hard to build a profitable server platform that can compete with what's already there. Engines are already very advanced, especially on the open-source front. Free resources are plentiful. Bookshelves are already crowded and chess-related articles probably don't make that much money. To make a successful app, you'd need some name recognition like Magnus. Coaching services are probably a no-go as well, since a professional player interested in that can always get a lucrative deal with chess.com or chess24 if they are high-profile or simply quickly get clients on their own and not have to pay a percentage.

All that is really left is targeting people who can't or won't perform even the most cursory research relating to their purchases.

Chessbase still has the most popular database software, which is used by many (most?) good players. That's not really an expanding market though.

I get the impression every good player uses it. And I've never heard anyone talk about using anything else in its place. People use it with the Mega Database, which ChessBase also sells, 8+ million games, including 85+ thousand annotated games. Just updating the Megabase looks like it costs 70 euros a year.

https://en.chessbase.com/post/new-mega-database-2021

It's happened before. https://en.wikipedia.org/wiki/BusyBox#GPL_lawsuits

https://en.wikipedia.org/wiki/Jacobsen_v._Katzer

involved GPL code.

> Everyone is allowed to copy Stockfish/Leela and sell them, provided the terms of the Stockfish/Leela license are met. But don’t pretend that the product being sold is something it isn’t.

What's the issue? Did OS contributors suddenly realize you can make money off software? They're upset at marketers doing marketing?

Don't understand the downvotes. If they violated the terms of the license or did something illegal then sue, otherwise they're just whinging.

If a seller only gets sales because of the ignorance of their customers, it's a rip off. I'd be annoyed if my work was being used to rip people off, even if no licence terms or laws were being broken.

Sounds like SOP for American business to me.

“It is sad to see claims of innovation where there has been none, and claims of improvement in an engine that is weaker than its open-source origins. It is also sad to see people appropriating the open-source work and effort of others and claiming it as their own.”

These seem to be their main gripes

Sounds like a standard software company...

They’re just letting the community know that they shouldn’t fall for the scam.

Apart from the question of legality of removing copyright notices/changing authorship (which is definitely a copyright violation in some jurisdictions), there is more than just legality.

One can (and most people do) condemn some actions even if they are technically legal.

It's a scam (it doesn't even use the best version of stockfish, so much for "the best chess engine available").

Moreover, it infringe GPL. I would love if a country started enforcing GPL with huge fines. Company-destroying fines.

The only people who can enforce the GPL are the people who own the copyright of the code infringed. They may be able to win huge fines, but (with some infamous exceptions) generally aim more for compliance than punishment.

>It's a scam (it doesn't even use the best version of stockfish, so much for "the best chess engine available").

Have you used the internet or turned on a Tv in the last 20 years? Do you really think Dodge has "The Best Truck in the world" or that Verizon really has the "fastest 5g in America". This is marketing 101 for any business in the US.

Meh, it's within the guidelines of the license. If you don't want companies taking your work and selling it, please use a different license.

Arguably no, taking the neural network weights into account. The Chessbase folks seem to think that this isn't part of the code, the Stockfish folks disagree.

As described in the article, selling it isn't the problem. Selling it and refusing to live up to the license obligations is.

That's up for debate, let the lawyers duke it out. To me, it's like suing for an xml file with different values. It'll be who has the bigger lawyers.

To stockfish dev, you should change your license.

This is a natural consequence of releasing an open source project. It’s the equivalent of leaving a car open with a key, documents and an invitation to drive it for free.

It's closer to releasing a genetically engineered racing driver for that car.

Anyone else is free to use said driver, but in this case if they make him faster they have to make their changes public because of the GPL

> in this case if they make him faster they have to make their changes public because of the GPL

Related to this: the GPL's obligations about sharing the source, kick in if you're sharing the binary. It's not relevant whether you made any changes yourself. [0] For compliance, the lazy derivative product will have to make the source available. I wonder if they've done that.

[0] https://www.fsf.org/bulletin/2020/fall/why-providing-source-...

Reply on news.ycombinator.com