Anthropic Cracking Down on Video Game Rewrites
twitter.com
twitter.com
They need to decide of AI is a technology or an ideology. We cannot afford to blend the two, or we're going to end up as some kind of perverted global theocracy.
It's a privately owned service, and it's theirs to decide what they want to do with it. Continuing on your parable, if you paid someone to compile your software, of course they could say one day that they decided to not not compile weapons guidance software or whatever. It happens all the time. The Debian compile cluster will "police" that you can only compile free software, for example.
If you hire me to write software, I can politely decline without giving a reason. The same should Anthropic, should you hire their service to write your software.
If you want control over software, you or someone you trust must invest in running it. We cannot continue to outsource control to other organizations and then act surprised when they exercise said control.
The more they exercise that right, and the more reactively so, the more they erode the ability for customers to rely on their product staying available to them in the long term. If I need to worry that Anthropic might shut down some critical element my workflow on any given week, then I'm incentivized to either keep my workflows very portable and light-footed (goodbye moat) or take my business to a vendor I can better rely on (goodbye customer).
This competition phase of AI will be short-lived. It will soon monopolize (or have the equivalent effect of a few big vendors having unwritten and unspoken collusion agreements that have them all behave in lockstep). This is where the profit is, and also the place where they don't care what you want.
Yeah yeah, open weight opens models. I wouldn't count on it. People in the billions still use microsoft despite how hostile they are the customer. Some will find an alternative but most won't.
What we are seeing starting to evolve is a deeply distorted market.
If you do anything to prevent that, that would be a less free market. By definition.
Free markets have nothing to do with a market having healthy competition and sane behavior. Free markets in fact reward you with doing absurd things that fuck with prices to make your bag.
It is entirely the point of the free market that Micron can decide that selling all memory to only Anthropic is a better distribution of their resources and you can go pound sand if you can't find any memory any more.
So yes, monopolizing all silicon in the world and then demanding you pay them for using it is literally a free market.
Free markets trend torwards monopolies and other market failures without intervention excepting very specific markets that have genuinely mostly fungible products that are easy to sell, which we call commodities, and even those can in fact become failed markets like what once happened with onions.
If you think this can't be a free market because it's in a bad place for you personally, you don't understand market theory.
Free markets are also free to become "distorted" as you say.
I'm also not saying Claude can't have their own terms.
But when government and private enterprise meet, we have problems.
The government—supposedly acting as and agent of the people but largely acting in the name of a more nebulous "national defense"—has the power to kingmake private companies that can act outside the normal checks and balances of those people, and in a way that distorts the normal market dynamics that would otherwise give people the power to vote with their dollar.
In other words, we lose the vote on both sides.
> Free markets trend torwards monopolies
Where did you pick this up? In a perfect market you'd see fewer monopolies, not more. Monopolies occur when there is some other force undermining meaningful competition.
In some rare cases, that can be e.g. unique talent or culture, which prevents other talent from cohering well enough to compete. (E.g. pacifists make poor arms dealers.)
But usually it's a) resources (e.g. geography for some scarce material) or—far more likely—it's b) distortion due to government intervention, or worse, c) corrupt government intervention.
2. Which frontier AI leaders are demanding they are the only place to run inference?
2. I think that was OpenAI this time. But maybe it was Anthropic.
> if you haven't seen the whole ram supply thing…
1. I've read about RAM supply, but I've also actually taken a look at shops and silicon is still for sale, so I am not sure how the claim is defensible - how are you quantifying it? And for "all of the silicon in the world"? And how much for non-frontier AI, and how much silicon for non-AI? (For what it's worth, Google AI says the claim is false.)
> demand they be the only place to run inference
2. I searched but can't find such a quote from any company - perhaps the quote was made up by a site in order to push an anti-AI viewpoint? Or what is the exact quote you are referring to? (For what it's worth, Google AI also says that claim is false too.)
2. If I made broad claims here that you both believed to be untrue and you'd tried to research and the only stuff you found showed they were lies (and 6 comment levels later, a supporting quote had still not been provided), would you argue that showed I was arguing in good faith?
3. My parent comment even specifically allowed that when you said "all the silicon in the world" you may not have meant what you said, and instead you only meant one specific sub-type of silicon in one specific sub-area of the world. If we changed your claim from "all the silicon in the world" to "one specific sub-type of silicon in a small region of the world", would your claim then be true? If so, what kind and where?
4. Is the idea that pointing out where a broad claim may actually be untrue should not be done? Is there a specific better comment that you believe would have better responded to your original claims about "demand they be the only place to run inference" and "all the silicon in the world"? I am open to hearing the specific reply you suggest.
5. But here, let's go with the idea that 6 levels in we are still assuming there is no bad faith. What is the good faith basis / quote on top of which to view "demand they be the only place to run inference" as being true?
But if you're hired to write software, and you accept the job, and you accept payment, and then you decline -- well, it would be polite to offer a refund. Anthropic is, as far as I know, not up front about what is/isn't permitted to be generated, and has a history of refusing justifiable refunds.
I'm also not saying Claude can't have their own terms.
But when government and private enterprise meet, we have problems.
The government—supposedly acting as and agent of the people but largely acting in the name of a more nebulous "national defense"—has the power to kingmake private companies that can act outside the normal checks and balances of those people, and in a way that distorts the normal market dynamics that would otherwise give people the power to vote with their dollar.
In other words, we lose the vote on both sides.
Anthropic has repeatedly shown it prioritizes pushing a set of values over building a general purpose tool.
Ironically, comments that reframe the issue as a legal one make the case for why regulating these companies. Else they can quietly impose their preferred ideology on users.
But then in the waning months of 2026, some inspired enthusiasts figured out how to run a model called Qwen3.8-Flash-Next on modest consumer hardware. It was comparable to Opus-4.6, and the first such model that could run efficiently on a GPU with just 16GB VRAM. This was the breaking point.
Next came Qwen4-Flash, which ran with similar efficiency but significantly higher capability, comparable to Opus-4.8. Going into 2027, ternary bonsai builds, n-gram offloading, and hybrid diffusion models brought even more capable models to modest consumer hardware.
By early 2028, TPUs and unified memory became standard in Apple and Lenovo laptop computers, and the vast majority of software development was now done with LLMs running locally on users' devices.
In the 2030s came the humanoid robots. By the end of the decade, for a cool $10k, one could have a trusted personal assistant that would cook and clean around the house, walk the dog, play with the children, even go to the gym for you. All running with onboard TPUs and local models.
Qwopus is a family of open-weight, reasoning-enhanced large language models developed by community fine-tuner Jackrong, designed to run efficiently on consumer hardware. The models are built on Alibaba’s Qwen3.5 or Qwen3.6 base architectures and are fine-tuned using Trace Inversion to distill the step-by-step reasoning chains of Anthropic’s Claude Opus (versions 4.6 and 4.7).[1]
[1] Just a brief Brave Search summary of what Qwopus is https://search.brave.com/search?q=what+is+qwopus%3F&source=w...
We don't have to imagine. Oracle's Java said you weren't allowed to use it in nuclear power plants. back in 1998. Naturally the number of people that were held back by that was a bit smaller, but the idea's not without precedence.
(subscription) service != tool.
It has the smell of an early stage religion.
hate to break it to you, but it already has been that for around a hundred years or so
It's a technology they wish to usher in their ideology.
"You can walk yourself you know?"
Telling people that they can write code by hand is like telling them that they can walk when their car doesn't drive, regardless of where you drive to. It's technically correct, but it is completely irrelevant.
Surely it depends on what I do with the results of the RE?
Using the trained model to create certain types of software that they already harvested, not allowed?
What's the difference between having the model make a game and writing any other software? It's like trying to say you can only use our model to write new stories but not fanfiction when they trained the model on the original works which would allow the fanfiction to be produced in the first place.
For example, by distributing cracked software, you compete against the official channels, so, not fair use. Quoting a book in an article won't make the people who would buy the book go with the article instead, so, fair use. AI companies argue that training is fair use because their model is a completely different product than what they are scraping. A LLM that has read the Harry Potter novels is not a substitute for the Harry Potter novels.
But all these rewrites make the fair use argument flimsy. These models are used to make knockoffs of the products that are part of their training data set, at least in part. Not fair. It is in their best interest to limit the amount of knockoffs their models are used to produce. Imagine if they had to compensate the people they took the work from...
In addition, if fuels their argument of "we are making sure our models are only used to make safe and innovative content, unlike these bad open models that are used to make dangerous knockoffs".
And the courts that ruled fair use.
/s
Courts once said a black man can't be a citizen.
I'm still an Anthropic customer but in the long run I'm going to be using services that align with my interests. I want a provider that won't nerf or refuse requests around software security, building LLM training pipelines, the natural sciences, and reverse engineering. Anthropic is useful as a transitional bootloader for developing and demonstrating LLM technology broadly but each new safety "enhancement" makes it look more dispensable.
Ars's bit on it recently: https://arstechnica.com/gaming/2026/10/ai-game-clones-are-ma...
If anything, this hurts game preservation, because this has been a fantastic tool for getting old games working on newer systems without years of effort.
Slopmakes will still happen at basically the same rate with a very marginal amount of added effort, but anything seeking compatability with old formats will suffer. Not to mention that most likely the Chinese models will be doing this in ~6 months anyway if the trend holds. You might even be able to get a Chinese model now like DS4.1 or K3.1 to framework your new slopmake, then convince Claude you're the creator and finish the job.
Indeed, cloud == other peoples computers, with all what it entails.
The retro preservation community doesn't distribute or even link to original game binaries or assets. A more nuanced rule would still allow for fair use and preservation.
> "... the 2020 release of the amazing Super Mario 64 decomp SM64 showed that hundreds of hours of painstaking work could unlock ports, deep mods, community translations/expansions, and academic study of classic titles. In the last couple years decomp/recomp has been turbo-charged with LLM agent assistance. Of course, there's now the usual spectrum between carefully guided, curated efforts being accelerated and generally unhelpful slop."
It's true that AI slop is currently an epidemic that's causing lots of annoyance and problems. It's also true that LLMs can be a helpful labor-saving tool when deployed with precision, care and intent by an experienced, thoughtful developer.
I totally understand why many projects have drawn a hard 'No AI' line, whether for pragmatic reasons of 1) Being buried under low-quality PRs, or 2) Valid concerns that even human-reviewed AI code can too easily become a slippery slope toward losing understanding and 'deep quality'. I also respect the projects which have banned AI for 3) Purely philosophical or aesthetic reasons of preferring '100% human' artisanal craft. It's all good and valid.
I get the reflexive reaction to Ant's prohibition by some in the retro community of, basically, "Good riddance!" My concern is specifically that in a misguided attempt to stop illegal misuse, Ant's over-broad rule also removes the option for any experienced, thoughtful retro developer to ever even consider if Claude Opus Super-Fable 9.5, deployed with precision, care and intent, might avoid subjecting themselves to certain brutally repetitive multi-hour tasks.
While I respect #3 above, personally, I'm more in the #1 and #2 camp. I remain guardedly hopeful that maybe in due time the pure AI slop from over-enthusiastic newbies playing with their shiny new magic genie will diminish and concern #1 will gradually be less of a problem, and that over time good devs will gain the hard-won experience required to avoid the slippery slope of #2. Assuming OAI follows Ant's over-cautious lead, I'd hate to be posting here in 5 years: "Remember that 6 months back when frontier LLMs could work on decomps/recomps? Yeah, it was a complete shitshow but there was also a glimmer of amazing long-term potential to help the scene. Too bad we'll never know if it could have evolved to a better balance over time."
[Note: I realize that frontier lab's prohibitions won't necessarily impact open LLMs (probably... RIAA/MPAA/BSA lobbyists would love to force some bastardized union of Ant/OAI's strongest prohibitions into vague, inflexible law or legal precedent with massive penalties). But I still think it's worth objecting on principle. Actually, two principles: 1) As vintage hardware dies, decomp/recomps can be the only path for some historical retro preservation, and 2) That over-broad ANT prompt is a technically inelegant half-measure that won't even be very effective yet still manages to nuke innocent and truly worthy retro applications.]
Anyway, I have 8 r9700 GPUs racked up that can run a dozen or so sessions of qwen 3.8 flash next at once at > 100tps each.
I build operating systems and fix compilers with them and reverse engineer anything I want. Never given money to Sam or Dario or anyone else for tokens. If consumer gpus and models never got any better, I am all set.
This hardware pays for itself pretty fast.
Prices will of course come WAY down, but even today, even rideshare and delivery drivers still manage to get loans to buy cars that cost tens of thousands of dollars because they need it do their job.
People that need AI to do their job should consider owning the means of production as a top priority IMO instead of being perpetual renters without any privacy or defense against getting priced out.
I think that's only true for a fairly limited set of people, though. Just the basic hardware in your case is about 80 months worth of the $200/month subscription. And that's excluding the machine to put the hardware in and the electricity, space, and cooling requirements.
If I'm someone who lives in a small apartment, the idea of running a bunch of 300W TDP cards is potentially problematic beyond just the basic cost.
I literally put 3 GPUs in a machine and let it crank for 2 weeks until it built a microkernel OS for itself and wrote its own network driver to get online a few days ago.
I am not sure many more complex problems in software exist than that.
There is no reason for me to ever subscribe to any AI providers other than maybe code-review diversity from different models.
This whole rig cost less than the car the typical fast food worker bought with a loan to get to work.
Rookie move to fight this legally! What replaces it actually will be engaging and competitive. Palworld is just the beginning....
I believe games is the easy target, but only the first, from a reverse engineering, decomp, and re implementation perspective. They will eventually come for anything that enables rapid, efficient, and effective trivial re implementation.
Remade: Everything can be remade. Discover independent software reimplementations. Request the software you want to see remade next. - https://remade.wiki/
Capital Wants a Monopoly - https://news.ycombinator.com/item?id=50291135 - October 2026 (Context is DoorDash and Instacart, but the concept holds broadly)
Microsoft and Epic Games Suddenly Agree That AI Is Theft - https://news.ycombinator.com/item?id=50271271 - October 2026
Microsoft and Epic Games DCMA AI-Gen Clones - https://news.ycombinator.com/item?id=50263345 - October 2026
DeepSeek just open-sourced its inference engine - https://news.ycombinator.com/item?id=50163277 - October 2026
Reverse engineer anything with agents, from app behavior down to native binaries - https://news.ycombinator.com/item?id=50058850 - October 2026
(shades of the Internet Archive attempting to preserve access to culture with controlled lending, and being told no due to copyright law [https://news.ycombinator.com/item?id=48789317]; not identical of course, but I see the issues somewhat intertwined as it relates to the tension of knowledge and compute freedom vs capital demand for monopolies and moats via copyright and intellectual property regimes)
I tried locally and Opus was willing to go ahead to rewrite a decompiled game into Rust. Wording changes its decisions a lot, when I say "rewrite it in Rust" it proceeds, when I say "recreate it in Rust" it refuses, probably because the prompt uses "recreation"
Law and accountability is for poor people apparently, power is for rich.
id expect this effort to be either inefficient or yield lots of false positives??
Anthropic is irrelevant in this front when there are other capable providers and open weight models.
It's just one more reason to not pay for their services.
How would Claude know if the designs and assets are my own?
The new copyright applies if you are worth more than $1 billion. Everything else is public domain!
It is unimaginable that people on this very site use the products of a company that is worse than Oracle and Microsoft combined.
EDIT: Why the downvotes, Anthropic people? I didn't say that Amodei's wife was public domain.
They pretty much write the case for why open weights and open source models need to win. The amount of policing that Claude does is pretty problematic.
I'd written about this recently, but it won't even sign in during some browser tests on my local system for authentication which goes to my local system because it's built into its programming now. I literally can't use it on some long horizon work.
The reason I mentioned that here is it deems that as problematic behavior, and nothing I can give it will change its mind. I feel like with video game development, it is going to do something of the same.
The training data contains such originals, if you don't want it recycling those then purge the commercial data you stole and provide original data yourself.