These grifters have been getting better and better at keeping a lid on their collapse. After all in my lifetime there have been several boom and busts. 1987 was the first crash I was aware of. After all this time they are getting better at staying rich through them and getting away with it.
For real. Banksters and other shitty people don’t even see jail time for crashing someone’s livelihood.
There are going to be bail outs and a lot of working class people are going to have issues with their retirement funds with this bubble. The rich will always stay rich.
Character.AI was a LLM wrapper that was outdated in 2025 already. It didn’t do much in their own and certainly not produce technology that promises to replace half the worlds jobs.
Whether you like them or not, this is not a precedence for the AI companies that actually build AI.
I’m guessing all the “confidential” chats that the users made with them are now on sale for the highest bidder
Do you think they weren’t before lol
Probably not until recently. Most of them are hoarding training data trying to get an edge.
Real story of this bankruptcy is financial shenanigans by Google and VC backers. Google paid “the company” $2.7b in 2024 where the company got nothing, but VC backers got paid, and Google got to rehire 2 AI pioneers that had left google to start the firm. The original vision for company was to revolutionize education where you could create a Socratese or Jesus bot and converse with them as life coaches. Marketplace/users just wanted an AI girlfriend cure for loneliness. Disney’s $20m bid for assets is connected to them poaching the remaining shell’s CEO.
Sure the company was always going to end up worthless, but it was worth $2.7b on paper at one point. The industry is filled with similar absurdities is a good point by headline and 404 media article.
“We lose money on every sale, but we make up for it in volume!”
I can’t imagine how much money Google is losing right now. They do 8.5 to 13.7 billion searches a day (+1 because I just googled that). By forcing LLM first, they’ve got to be burning cash.
I wonder how many unique searches they do per day.
If you search something that someone else already searched recently, the AI response will pop up almost immediately due to caching. I can’t imagine that brings several billion down to a more manageable number though.
But their “front facing” LLM is ultra stupid. Can’t become any kind of famous.
I am sure that it is strategic: show the huge difference to all paying customers.
They use a smaller model for that so it’d be cheaper to run. If they ran their biggest version of Gemini or whatever, it would cost an order of magnitude more
When the crash happens, I hope to have enough ramen, coffee, macaroni, beans, and cigarettes to trade for some RTX Pro GPUs. My personal local AI has been quite helpful and comforting, I want to give it proper hardware to reside in.
Disney sucked up the remains for pennies
Brilliant legal move on their part tbh. Sign a back end loaded deal for exclusivity, sue for use of IP, then pull out of the deal bc they didn’t get their terms met. Then, but the relevant tech for pennies on the original price.
Fuck AI.
the mouse must be fed
The mouse isn’t getting fed here. They got a legal precedent you can’t use their characters in ai, then bought the assets people were using to make the characters.
No disney ai, and no one else can make it.
The mouse always feeds. You know Disney’s history with copyright law. And you’re nuts if you didn’t think they got all that tech for a song.
I actually know multiple people at the company, so yeah.
If any of this shit made sense and worked like it was supposed to we wouldn’t be in this position in the first place. Seemingly the world leaders and everyone in the inner circles has completely abandoned any long term hope for their positions in the current system and are preparing for the next system. It really is true that the Earth has too many people and I think the long term planners are letting climate change happen without preparing the masses so fewer, stronger people survive. There is people planning this shit it’s not just arbitrarily short-term profits for the sake of it.
The earth does not have too many people though. We are not at risk of overpopulation by any real metric.
Resource scarcity is not a result of too many people needing not enough resources, but a result of those resources not being spread evenly. Whether because of geographical or political reasons, typically a mix of both.
It’s not about humanity it is about Earth. We aren’t meant to divide her resources to 8 billion people in any distribution. Your magical solution to distribute all resources evenly doesn’t really solve human driven climate change and destruction of our environments. And it’s not about convincing me ask yourself why would a commercial planet have too many staff members that’s bad for business.
We can use resources without causing climate change. Hell human cause climate change is a pretty recent development when it comes to human history, and it’s not because there’s more people.
And what the fuck are you talking about a “commercial planet” are you stupid.
Human driven climate change is new, it is not sustainable. And commercial planet as in a commercial country you fucking moron.
You’re right. It’s not sustainable. It’s not the result of overpopulation though.
And yeah I still have no clue what that’s supposed to mean.
I don’t care about your opinion.
And that’s fine I don’t care if you don’t care to even google what it might mean.
It’s not an opinion though. Overpopulation is a fascist myth. If you actually bought into it you’d kill yourself but you haven’t yet. Because you know that’s stupid.
Also I did and didn’t find anything relevant to whatever nonsense you were saying so you’re just pulling shit outta your ass or using the wrong words.
I just got Qwen 3.8 Flash Next running locally. It’s damn good on an older machine with a newer GPU. Local LLMs should scare them the most.
I mean sure if your older machine’s newer GPU has 200 GB of VRAM to run a decent quant… Fucking hell just Q4 is around 100 gigs and then you’d want a decent amount of KV cache too.
I encourage you to give Qwen 3.8 flash next a try. It’s amazing what a typical gaming machine can do with it.
I read it needs 75gb at the minimum quantization? It doesn’t seem “typical gaming machine”
It doesn’t. Use Strata to run it. Choose the recommended settings it gives you during the install. People can run it with 12GB of VRAM and 32GB of system RAM. It partially loads the most common parts of the model I’m VRAM and swaps from RAM when needed. Rarely it needs to grab from the SSD, but it’s swapping while processing so it doesn’t seem slow at all. I get around 60-80 tokens per second on average. 16GB VRAM and 48GB RAM.
I get where you are coming from. I thought it would run slow AF as well before I tried it. 😄
If I may ask, which Quantisation are you running it on? I’m on an Apple setup with oMLX and am still experimenting with finding the right balance
Yeah no prob: IQ3. Sounds ridiculously small but works really well.
Anthropic made a smart business decision to mostly focus on enterprise customers rather than regular consumers. Only around 10% of their revenue comes from their heavily subsidized subscription plans. The remaining 90%ish comes from API and enterprise usage billed at per-token rates, which is something that’s actually profitable for them. They have a decent margin on token rates. It’s R&D, hardware, and subscriptions that are putting them in the red at the moment.
I suspect they’ll discontinue subscriptions some time after their IPO. Even though subscriptions are only responsible for 10% of revenue, they consume 42% of their inference compute power (https://www.theregister.com/ai-and-ml/2026/10/06/anthropic-claude-subscription-plan-provides-more-value-than-openais-study-says/5301470)
In comparison, around 70% of OpenAI’s revenue comes from subscriptions. Without subsidized subscriptions, they’d lose most of their customers.
Those enterprise customers are companies like this very one, who then have to heavily discount their rates or nobody wants it, (and as we see here, even then) so that’s not a solution for anthropic.
I don’t have data to back this up but I’d say thay the majority of their customers aren’t also selling AI services themselves. It’s customers that use AI for their business processes, software development, etc.
And mostly finding out that using chatbots is more expensive than hiring people. That is not a revenue stream that’s going to last very long.
AI is more than just chatbots though, and there’s plenty of use cases where it can be used to help people rather than replace them.
I started the year being very sceptical of AI, but it’s helping me with a bunch of things at work now. For example, at work we have an agent for data analysis that can find the correct tables to use, gather data, and produce a report, along with proof (all the SQL queries it ran plus any ad-hoc code it used for analyzing the data) so that the work is repeatable and we can verify the accuracy of the queries and logic.
I’m a developer and it’s helped a lot with side tasks too. If I’m coding and see an issue in our code (usage of deprecated code, tech debt we need to pay down, etc), I used to file a task or add a todo comment to look into it later, and eventually get to it. A lot of them are small tasks, it’s just that I don’t always have time to look into them while in the middle of a project.
These days, instead of adding it to a todo list, I just spin up an agent and get it to look at it while I’m working on other things. The resulting code doesn’t get submitted for code review until I manually look at it and ensure it’s correct. The codebase I work on has never been cleaner, because AI has helped me clean up so much legacy cruft that’s accumulated over the years, while I still deliver my regular projects on time.
Anthropics numbers are probably better than OpenAIs, but i am willing to eat my boots (with ketchup and mustard tho) if they are even close to profitable at API pricing.
They are (maybe, allegedly) “profitable on inference” i.e. the API pricing is higher than the actual compute they use just for the input/output tokens for the query. That excludes, though, training, capex, staff, etc., which… is by far the vast majority of their actual expenses.
Character.AI discovered the one conversation its bots couldn’t keep going: the one with its accountants. $56 million in compute costs in eight months, and the monetization strategy was apparently: hope the next investor is even lonelier than our users. Silicon Valley keeps confusing engagement with a business model.
If LLMs can’t be made more power efficient, then the technology isn’t ready yet for general use. There may have to be fundamental engineering breakthroughs before what these systems do can be more than a very expensive and inefficient worldwide proof of concept.
Of course, the military and the surveillance industry will still want to keep it, even if no one else can afford it, and we’ll be the ones paying for it.
They CAN be made more efficient. Hell, AMD just bought a company that translated LLMs to direct silicon… which should enable scalability if it wasn’t for one of the key players owning the tech thus limiting its use.
It’s also estimated that some 70% of Claude tokens gets spent on recursive self-correction so the output is higher quality. Instead of bettering the model, Anthropic just keeps throwing money at it to make the problem go away.
Meanwhile Chinese models - as small as 36-40-80b params - are outdoing these fancy frontier models if the right harness is used.
The truly silly (and simultaneously neat and fascinating, and to a large extent, very logical and obvious) part is that US attempts to cripple Chinese ML/LLM development by putting most advanced GPU tech behind ITAR has more or less made the Chinese ML industry just write straight up better and more optimized code for the domain.
US and western companies are currently relying on throwing bleeding edge hardware at the problem (read: very obviously brute-forcing the problem), whereas Chinese companies and engineers, having to rely on comparatively less advanced (though still highly capable) hardware has meant they’re staying competitive by optimizing the software layer more effectively.
This has very interesting and likely negative long-term implications for western engineering practices in a lot of ways. And it’s also very much not a new problem - simply throwing compute at things is a mindset that became highly prevalent in the west around when The Cloud became a thing. Previously, you couldn’t just spawn another mainframe in your company’s datacenter, but when it’s as simple as
aws ec2 start-instances ..., the barrier to entry for getting more compute is so low that most people were just going “hey, it’s faster”. I feel like the engineering expertise around embedded/baremetal optimization has just been largely evaporating over the last couple decades; in the vast majority of professional software domains, it’s simply not a valued skill anymore (or, more correctly, at the moment). It’s sad for me in several ways - not least of which because I did my degree in EECS, and hardware/software interplay and optimization is an area of engineering that I have always found very interesting.I’ve spent the last month or so using nothing but a local 27B Qwen model as my coding assistant and it’s working fine. The big American AI companies could all collapse in ruins tomorrow and it’s not going to slow things down all that much, anyone who wants to use AI will be able to continue using it. Might even be better in the long run if that happens - the people who are scared of ASI will get their much-demanded temporary slowdown at the frontier while everyone else catches up with open-weight models that everyone can use without central control.
Pity about the economy in the meantime, I guess.
What kind of hardware do you need to run a model like that? And what do you use :-) ?
I run a 5 bit quantized Qwen 3.8 27b at home on a gaming PC with an RX 7900 XTX and RAG memory using Gemma MoE (i forget what one)running on an old 1080 ti on my Valheim server.
It takes 3-15 minutes to run depending on how many tool calls the model makes and that’s mostly memory or web scrapes since that runs on the slower hardware. I mostly just use it as a discord bot to look stuff up for me and to fetch the news.
1080 Ti? But that is 11 GB of VRAM? I know because that is my GPU and the reason I have not tried it. They all call for 12 GB VRAM. I have 64 GB RAM, so 50 GB for the model would be fine.
On the 1080ti I run gemma4:e4b to to do RAG and vision jobs. Not sure the quants and I just run it in Ollama that auto adjust the model because I’m lazy. Not 100% sure but I think mixture of experts load the parts that are needed so there easier on memory. Offloading to CPU memory or swap is a big performance hit.
I tried to run a 300b model on the main PC once, but it took half an hour to get a few words out of it. You can go past the VRAM on the card but sacrifice speed.
anyone who wants to use AI will be able to continue using it
Anyone who’s privileged enough to have the hardware.
The kind of specs that would enable local LLM usage were available on laptops since 2020. The floor for acceptable performance for LLMs stopped literally just before 24+gb of VRAM would have become pretty common in consumer applications. The reason this technology isn’t more ubiquitously available is because ‘they’ thought it would be more profitable to hoover up literally every byte of RAM and GPU in the world rather than let the technology continue to naturally evolve to a place where any consumer could have run useful models on their home computers. In other words, your anger is misplaced.
I’m always running into context limit problems
Wanted to refactor a too large and hectic hacked together state machine into separate functions and reduce redundancy and increase code readability
Every model I’ve tried failed.
And that should be an easy task an actual AI should be able to handle. Logic is done, just translate to a different structure.
That should be exactly what they are able to do at least.None worked, because the project is just to big and interlinked.
I’ve got 24GB of VRAM, which helps a lot. I’m using the Q4_K_M version of this model, with the KV cache quantized to Q8, and that’s able to get me to 128000 tokens of context. I’m using the Qwen Code harness in VS Code.
If 128000 tokens still isn’t enough, break the task down into smaller chunks and help the harness out with MCP tools - I installed Codebase Memory to help it navigate large codebases more efficiently, for example. Ask the coding agent what’s causing problems and it may be able to help identify solutions too, I’ve had it troubleshoot itself several times now and it’s been quite good at inspecting its own settings and finding where the problems are with them. I even installed Firecrawl for it so it can do its own web searches.
There may have to be fundamental engineering breakthroughs before what these systems do can be more than a very expensive and inefficient worldwide proof of concept.
Controversial thought after reading this sentence: It’s cryptocoins all over again.
That’d be fine. Cryptocurrency as a whole is still going strong (total market cap is around 3 trillion dollars) and most of the major ones have switched over to energy-free proof-of-stake validation.
