Robust Mirror

They certainly don’t want people running their own models but it’s a stretch to say they orchestrated it rather than it being a side effect of what they were doing anyway.

In any case I’m largely referring to what we have been told by independent researchers. Yes anyone in the field should be taken with a grain of salt but it is far more ridiculous to make entire conclusions based on nothing but vibes and a dislike of AI.

They don’t own an AI model, and there are objectively 2 different models owned by 2 different companies in that article. I have no idea what you think you are reading.

If you aren’t even going to attempt to read things I don’t know how to help you much more. Them selling something doesn’t make the research automatically wrong.

I’m done with this conversation now. I’m fine with people disagreeing with me but you are not even trying to discuss in good faith.

https://newsletter.semianalysis.com/p/anthropic-subscriptions-offer-5x

Yes there is such evidence. You can argue all day about whether the evidence is up to your standard, but it does exist. Remember that evidence does not have to be hard, absolute facts directly from a primary source. It can be inferences based on the information we have available to us.

When deciding whether someone committed a crime, for example, we do not need them to tell us directly or have footage of them doing it to have evidence we can weigh.

Sure, but assuming we can trust what we have been told so far. Regardless IPO is very soon, there they can’t easily lie or hide anything. If the first part is meant to be “see, you agree”, my point is the amount they need to change by isn’t necessarily as big as some might think.

In any case what happens is what happens, I just like to weigh as much evidence as possible when looking at these things before making assumptions.

It’s not as bad as you think. Anthropic is running at a loss at max usage. Openai isn’t. And even Anthropic profits if you use 20% of your quota or less each month. They would only have to move the price up if it became a real issue. Or cut usage cap down (that’s mostly how openai does it, $20 gets you less total tokens, there’s nuance to whether this is automatically worse though). But not as dramatically as you might think. It would lock out many hobbyists, but not companies that have come to rely on it.

Not posting this to defend them but because the nuance matters to understanding what is happening better.

The $80.7 billion loss looks huge, but most of it isn’t straight cash, about $72 billion is a paper charge. Anthropic issued bonds that can turn into shares. This means they get more valuable as the company does, and Anthropic has to count that rise as a loss. So it’s actually a case of the loss being “higher” the more the company is valued.

The actual reported operating loss in all 2025 was around $8.1 billion on $4.6 billion of revenue, with Anthropic saying it’s now profitable on an adjusted basis, though the adjustment is their own way of counting and what will be under scrutiny.

I’m not saying it is the best ever made but I bought this 5 years ago and still use it for TV, movies and gaming in the bedroom and am very happy with it. No smart features.

ΟΡΤΟΜΑ GT1080HDR

If you’re still interested try other models like Deepseek, Kimi or glm, they were all willing to engage with it when I just tested asking about hacking LG 65" QNED70BSA AI Mini-LED Smart TV.

From what they said LG is patching aggressively though. The most useful resource I got from it is this site: cani.rootmy.tv

QNED70BSA Found QNED70BSA.AAU (webOS 26 (11.x), machine: k25lpn, otaId: HE_DTV_W26P_AFADATAA) Unrootable or unknown (yet) No known rooting methods are available for this model. Contact us to help us find a way to root! ©2025 RootMy.TV & webOS Homebrew

It will. But if you frame it as wage, it’s giving everyone a 1.5c/h pay rise to watch them. Many companies would take that. And that’s not considering the local models that are essentially free. Again I don’t think most are doing it.

And you’re right the paperwork and actually getting something built and installed and working across a company would be a massive hurdle. I’m just saying price isn’t really. Especially if they’re a company already paying for AI in some way.

I believe the vast majority aren’t, and I know money is money, but it’s not that expensive. Not only are there vision models small enough to run locally entirely free now on hardware more modest than you would think, even using an api key from open router and sending it a screenshot every 15 seconds to summarise it into a log costs in the realm of 1.5c/hour.

Now yes that is per person, but even so, 10 employees 15c/h, 100 for $1.50/h. And remember this is an unreasonably often every 15 seconds. Most would be happy per minute or even less. Under 40c/h to watch 100 people. It’s money yes, but it really isn’t that expensive.

Edit: voice is also cheaper than you would think to have a model stt and summarise it, also single digit cents per hour of audio.

Same in Australia regarding doubling down. They’re giving trolleys with an ipad thing on them you scan everything with, and increasing the exit area security so you cannot leave without the staff opening it if you didn’t buy anything, and AI based cameras all over the store and gates that judge whether it should open for you after buying things.

It’s a mix of awesome and horrible.