It’s out of the box.
They can try, but, it’s too late.
Don’t hesitate, bring them to thier knees.
Is reversing with Deepseek just as good?
If not now, probably at some point it will.
No idea.
The legitimate developers I know aren’t hit by Claude restrictions, because they don’t fucking use it.
Hilarious
“Give us all your code for free who cares about copyright anyway”
“NOOOO YOU CANT USE OUR PRODUCT TO GET FRE CODE 😡😡😡”
AI exists to aid capital against you; not aid you against capital.
I suspect we’ll see these features opened up for US-allied actors against their foreign competitors.
You can guarantee US auto companies are going to be free to reverse engineer software and manufacturing techniques of Chinese companies, for instance.
But you’re not going to be allowed to use Grok or Meta to hack your Tesla or BMW, pull out the OS, clone it as freeware, and offer a substitute unconstrained by subscription fees and gated upgrades.
Why just games? What about the Photoshop clone? Seems like making a derivative of anything should be blocked, but that’s all it can really do in the first place.
Seems like making a derivative of anything should be blocked
So we should get rid of anything in-the-likeness-of? Like Libre Office Writer being in the likeness of Microsoft Word.
Microsoft word was not the first.
I think their point is that it was trained on literally billions of repos and other software written, but because those developers don’t have hundreds of lawyers on retainer with an unlimited budget; their intellectual property rights don’t seem to be a concern to the slopmasters.
Major software companies and publishers are realising AI companies put their profit motive at risk.
Why buy the latest Mario game when you can have a version custom built for you? Why get an Adobe subscription when the cloned version is available?
That can be applied to any software now. Two industries are facing off.
This is why AI generated content should belong to EVERYONE. There should be no profit in it, because the content used belongs to humanity.
AI should belong in the history book of bad technologies, alongside with leaded fuel and asbestos.
I once asked Claude to check the app it vibecoded for security flaws and it didn’t want to do it.
It literally told me to look up some basic security principles and do it myself.
Yeah their anti-hacking protections are over eager. Ask it to find bugs instead, or reasons we shouldnt deploy yet.
I hook mine up to some classical static analysys tools like sonar or codacy. Just ask it to remove the security findings and haven’t had an issue. Then again, been on vacation while all this shit has happened, so don’t know what to expect when I return.
The restrictions are so they can gradually institute tiered access fees. This is how the industry becomes profitable.
Tier based models = economic segregation. The poor can fuck off and the rich can pay for enterprise.
Great time to know how to do things yourself.
The best part of paying for a coding assistant is discovering halfway through your project that it has developed opinions about what your project is allowed to be. Imagine your IDE refusing to compile because it thinks your game feels a little too much like Tetris. At least a compiler doesn’t pretend it’s doing you a favor.
That’s what happens when you let others do your computation for you.
Run the tool on your own hardware.
I think that’s what they wanted from the very beginning, to accelerate projects that are socially convenient for them, but not those that are not.
There’s that issue with universal machines that some people don’t like that others are allowed to make universal decisions.
I got shut down by it two days ago. It suddenly told me that the project I was working on had tripped some kind of guard, that it couldn’t continue working with me on it, and it couldn’t tell me why. I just went back to my local Qwen model. I can’t see this doing great things for selling Claude subscriptions.
I wonder if you can trip this on corporate subscriptions just by examining standard code so I could finally have an excuse to not use it at work.
My coworker recently had an experience where he was trying to fix a security hole and his AI wouldn’t do it because it claimed that it was unsafe.
This is another reason why I’ve reduced my dependency on AI considerably over the past few weeks, and I no longer personally pay for any AI tools/services at all.
It’s insane to think someone would pay for something and after using it for a couple days they get told they can’t use it for certain things anymore.
They enshittified too soon, before people depended on them.
Oh, based on my experience, there are plenty of people convinced they can’t do anything without Claude anymore. Also based on my experience, I would be ecstatic if they actually stopped trying to do work because they sucked both with and without Claude.
Maybe not enough to fund their financials, but that was going to be a tall order anyway.
That’s a pretty significant part of the value and utility of AI…I bet stealing ppls ideas and getting to market first is still totally fine though, right?
Local LLMs are the way to go on AI, they can’t suddenly change the model up on you. And also, I guess Anthropic just wants their models to be more useless ahead of going public? Because I presume people that want to use AI, who don’t care where they send their data to, but don’t want to nagged about how they can’t do benign programming and analysis tasks, will just choose another model provider.
They are the way to go, but hardware pricing makes it cost-prohibitive to match frontier capabilities, which makes decomp and recomp difficult. Or rather, they require more skill and take longer, introducing some barriers against the current floodgates being flung open.
We all know the eye-watering prices GPUs and RAM cost now, to say nothing of Unified Memory devices like the Strix. Most people just don’t have a spare $10K to throw at an inference box that can run (say) a 300B model at interactive (not batch) speeds.
Coming at this from another angle tho, I note that according to Steam stats, the most popular consumer GPUs are in the 8GB and 16GB range…which is smack bang in the VRAM range of what a good MoE (say, Qwen 3.6 35B-A3B) runs on (for 8GB) or arguably the best local dense coder (Qwen 3.8 27B) can live at. Which leaves us with a capable assistant and a capable coder.
I think there’s probably room for LLM improvement in the 8–16GB range too…and given that most consumer PCs and laptops live there, I’m sanguine the open-weight devs might make it happen.
But I also suspect the gatekeeping of high-end capacities is intentional.
We’re probably not going to have the ability to vibe code Mario into Elden Ring again…not without significantly more skill. Dunno.
Uhm yeah good luck with running anything locally after hardware prices double again and again. Big tech needs to be stopped. There is no wiggling out of this just by being a smart customer.
Thing is, I already am…
The price did grow several times in size, but even in the 4 to 5 digit price range, you or smaller companies can have a pretty competitive performing open source model as the frontier models from earlier this year and they are keeping pace.
Obviously I’d like the speculation bubble to end and the DRAM and GPU cartels broken up, as it’s late-stage capitalism where the existing regulations being unenforced on big tech because the regulators are in on it that’s the problem, not so much the AI tech itself. (Don’t get me wrong, the tech has bad sides to it too like stealing artwork, data centre environmental impact and whatnot, but just speaking to the cause of the component greedflation).
Yes it’s disgustingly expensive. I spent a whole lot of money on an old used GPU and it felt bad, but it didn’t feel dirty like paying for Claude would. And then a couple of weeks later the used prices for this GPU had risen by about 15%.
(I only use Claude at work when someone else is paying. And they’re about to make me redundant because Claude is cheaper than me.)
Is there a local LLM I can run on an 8-10th gen i5 with 16GB RAM?
Yes, quite a few, but it depends on what you’re trying to do with them. Any ideas?
Assist with writing letters, home assistant code, light powrshell scripting.
Edit: helping segment my network
That should be well within reach of a local LLM.
If you’ve got a discrete GPU with sufficient VRAM, something like Ornith running on PolyStrata could be worth a shot. Try a Q3 size.
https://huggingface.co/bartowski/Ornith-1.5-35B-A3B-GGUF
https://github.com/VecSzn/PolyStrata
Hell, Tiny Ling 3 might be able to do it. Try a Q6.
https://huggingface.co/bartowski/Ling-3.0-tiny-GGUF
Another good option is Qwen3.5-9B. I’d try a Q4 or Q5.
Whichever model you choose, I’d also suggest a few MCP servers to give it access to reliable information rather than rely entirely on its internal knowledge.
- Microsoft Learn MCP: Official PowerShell, Windows and network documentation. Will help with your scripting stuff.
https://learn.microsoft.com/en-us/training/support/mcp
- Exa MCP: Web search for current technical information, Home Assistant documentation and code examples.
https://github.com/exa-labs/exa-mcp-server
- Filesystem MCP: Allows the model to examine your existing scripts and configuration files rather than guess what you’ve got.
https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem
If you’re actually running Home Assistant, its official MCP server is worth a look too.
https://www.home-assistant.io/integrations/mcp_server/
I’d start with Microsoft Learn and Exa, then add the others as needed. You’ll need an MCP-capable client (which pretty much all of them are now - eg: llama.cpp web ui, OpenWebUI, Light Local Chat etc), and I’d keep file access read-only initially.
I’d personally use McProxy instead of adding MCPs singularly, but I fear I’m already over loading you with information.
EDIT: The PolyStrata recommendation assumes you have a discrete GPU with sufficient VRAM (unclear if you do?). If not, disregard Ornith and try Tiny Ling 3 Q6 or Qwen3.5-9B with ik_llama.cpp instead.
https://github.com/ikawrakow/ik_llama.cpp
Either way, both smaller models should be capable of the tasks you’ve described, particularly with access to the right documentation. Just verify any network or firewall changes before applying them.
With an iGPU results are going to be very slow. Most promising things I have heard of so far is Mixture of Experts type models, if you have an 8GB card you might be able to run Qwen 3.8-35B-A3B-MoE which I think is the most capable for that hardware.
There are websites that can suggest a few for your specific hardware, I assume some are vibecoded and are only occasionally updated to the latest models, so take it as you will. See this Lemming’s comment. And !localLLaMa@sh.itjust.works is where I think most of the discussion around new development with Local LLMs happens here on Lemmy.
They suck far too much at the moment.
I… Completely agree and my comment wasn’t meant for you. You have no idea how often I encounter people that simply do not understand or know exactly what these llms do and they simply dislike them for the sake of being part of the bandwagon, I am generally irritated by these people. The ones that say out loud that they hate everything that AI does and then when it does something generally useful and they don’t hate that one particular thing. It’s extremely two-faced and inconsistent and absolutely bothers me.
you’re on the wrong platform buddy.
go run off to Reddit where the bots will agree with whatever bullshit you’re spewing.
They’re training their models on my art. It’s only fair I get to use their technology in return.
