Legal risk from Microsoft and other game makers. Path to success will have to be on prem/self hosted LLMs. Big Tech wants a monopoly on AI revenue, they do not want people cloning what they're gating with IP for enterprise value and cashflows.
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/
IP protection once again scales with your wallet, where the benefits disproportionately help those who already have the most resources and hinders those with the least.
They didn't start with games. At least they had plenty of arbitrary limits for what it wouldn't do usually citing safety. What they do now is introducing a new arbitrary reason - this time because it hurts other deep pocketed buddies, large companies in the gaming industry. You can bet your tokens this isn't where the list of arbitrary reasons ends.
I'm not sure why it's being limited to just games (it might just be the announcement), but I'm sure vibe-coded stuff like this that people keep happily plastering over every possible social media outlet presumably isn't helping.
That's not even a recreation, it's just the leaked sources of the game recompiled to WASM, so I don't think their prompt would apply to this case.
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"
Hard to imagine this is actually a concern for game makers. Decompilation aside, AI can only produce shallow visual mimicry. If you were to actually try to enjoy these games you'd be miserable nearly instantly.
Rookie move to fight this legally! What replaces it actually will be engaging and competitive. Palworld is just the beginning....
Can you imagine a devtools company deciding, "Actually, no, we're going to police what kind of programs you can compile" ?
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 the natural consequence of a few companies having all of the compute resources. They can do whatever they want since the alternative for users is to pay out the nose for hardware to run a model that isn't SOTA.
My great hope is that we will look back on this as a tiny sliver of time, when everyone couldn't resist the awesome power that came with running our devtools over the internet through megacorps, willingly handing over our code and secrets and dollars, in exchange for those sweet lines of code.
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.
We have models right now such as Qwopus, which are Qwen with open weights and using distilled Claude responses in it's training data. Since AI can't be copywritten, distillation is the way forward to getting rid of these proprietary 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]
So people will be running reverse engineering through openrouter and similar or local inference while Anthropic is severely crippling cyber security usefulness of their products. Feels like own goal.
Anthropic (and others) mass harvested source code from around the internet while ignoring licenses to build their product. That's of course allowed (by them).
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.
Anthropic and others harvest source code from around the internet based on the idea of fair use. The big idea behind fair use is that you are not using the copyrighted material to compete against its owner.
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".
So we are going from running your stuff from cloud that randomly closes accounts or charges ridiculous costs to pinpointing arbitrary AI use cases and banning them on a whim. I guess that's progress !?
Indeed, cloud == other peoples computers, with all what it entails.
Still blows me away that we have 20 years of history showing that this is always the end result and yet people are still surprised when it happens. The memory retention of tech people is like that of a fruit fly.
Mark my words: AI will only be allowed to destroy the software engineering profession, but somehow protectionism will be enforced for everything else in the name of “safety” and “fairness”.
All those unemployed SWEs will just have to work together and build plausibly deniable local AI tools that can be passed around through covert channels.
Easy to see the slippery slope from "safety" to "DRM". Abliterated/open models will become increasingly important for AI freedom if they aren't made illegal first...
Same way you would with missile guidance/targeting software (ITAR in the US, or equivalent in other MTCR countries). You probably wouldn't prosecute individuals running models locally but create a chilling effect by scaring any semi-upstanding citizen from openly hosting/developing models (and no US business would touch it with a 10ft pole).
The most important secret of the atomic bomb was that it could be built at all. That's the same way I feel about Anthropic's products (increasingly, tech demos rather than products, if you go off the happy path.) They've demonstrated what is possible. Now I'm hoping for a fistful of rivals to build functional equivalents.
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 them look more dispensable.
They can't help themselves but play the Don't be evil[0] in fast-forward.
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.
How can something that provably cannot create and only recycle follow such instructions.
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.
I wonder how long before people figure out that you can run alliterated models locally on consumer hardware capable of doing all this without help from Sam or Dario.
Damn spell check. I typed it correctly the first time.
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.
But yet they can afford to pay Anthropic $200/mo forever?
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.
Those are all pcie cards marketed and sold to consumers that run on consumer boards. I bought one at a time over months whenever I find cheap used ones on ebay, always under 2k. A single card would be plenty if this was not my career and if I was not working on many things in parallel.
This whole rig cost less than the car the typical fast food worker bought with a loan to get to work.
I very much want this to be true, I would shift as much of my workload as possible to local models. But is qwen 3.8 really good enough to do reverse engineering work? Is it because you have many sessions running at once, so it can "brute force" its way through problems?
Many sessions working in parallel breaking a problem into small parts can chew through any complex task I have been able to imagine. Even complex software reproducibility and compiler bugs that previously I spent months solving as one of something like a dozen people in the world with substantial expertise in this area.
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.
Some context to this, there's been a bunch of discussion in indie game development spaces about how it might be risky to release demos, or any information really, since a bunch of (probably? can't necessarily assume now.) people seem to immediately try to clone it with agents. It's to the extent I've seen genuine confusion about which game is the original in Twitch streams/chats. There have been a bunch of takes one way or another on it, including 'if your game is clone-able from a demo your idea isn't unique or complex enough'.
I wouldn't even say this fixes that problem, it just adds a slight barrier above pointing it at existing code and telling it to remake this thing. Someone could still thoroughly describe a game and get passable slop results not too far removed from the original. Unless the game is well-known, this probably isn't fixable.
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.
Anthropic is truly an altruistic company, selflessly working so hard to write free marketing for its competitors.
That one Anthropic blog, "hey check out how amazing GLM-5.3 is, wow", still lives rent-free inside my head. They're like the Gerald Ratners[0] of software marketing.
66 comments
[ 4.0 ms ] story [ 818 ms ] threadI 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
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
https://news.ycombinator.com/item?id=50259643
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"
Rookie move to fight this legally! What replaces it actually will be engaging and competitive. Palworld is just the beginning....
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.
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...
(subscription) service != tool.
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".
Indeed, cloud == other peoples computers, with all what it entails.
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.
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 them look more dispensable.
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.
[0] https://en.wikipedia.org/wiki/Don%27t_be_evil
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.
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.
This whole rig cost less than the car the typical fast food worker bought with a loan to get to work.
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.
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.
How would Claude know if the designs and assets are my own?
That one Anthropic blog, "hey check out how amazing GLM-5.3 is, wow", still lives rent-free inside my head. They're like the Gerald Ratners[0] of software marketing.
[0] https://en.wikipedia.org/wiki/Gerald_Ratner#Career