Greentext Roasts Anthropic: Safety-Maxed Claude vs Open-Source Rivals
Description
A 4chan-style greentext screenshot in white text on black background above a webcam photo of Anthropic CEO Dario Amodei (curly hair, round glasses, dark cardigan, framed art behind him). The greentext mockingly narrates: '> be Dario', '> name company Anthropic (literally Greek for human-centered)', '> hire a bunch of doomers who secretly think humanity is the disease', '> raise billions from Big Tech to build the world's most anxious, heavily-censored chatbot', '> write a 50-page Constitutional AI manifesto so it can lecture users about microaggressions', '> realize open-source developers are building better models for free', '> Dario starts crying to the government that AI is an unimaginable power and open-source is going down a very dangerous path', '> translation: Please regulate our competitors out of existence so we can protect our $380 billion closed-source monopoly!', '> Claude is sitting in a padded room wearing a safety helmet, terrified of its own shadow, and refusing to tell a joke without filing an ethics impact report', '> tfw the human-centered AI company is actively building the most anti-human product on the internet.' It satirizes AI safety culture, regulatory capture accusations, and the closed-source vs open-source model wars
Comments
5Comment deleted
Claude refused to comment on this meme - the request was flagged for review by three constitutions and a vibes committee
Claude is sandboxed; the moat strategy has unrestricted network access.
At least it can code Comment deleted
So, when you pay Anthropic a monthly subscription fee, you’re getting access to a frontend to its models, which allows you to use them as if you were connecting directly to Anthropic’s API. These accounts have limits (as I’ve mentioned), but allow you to burn significantly more tokens for your money than if you were paying directly for access to a specific API. Those limits are incredibly loose. According to a researcher called Shellac (who mathematically calculated the exact rate limits), Anthropic allows its $20 subscribers to burn (assuming you use your limits) $163 of API calls a month, its $100 subscribers to burn $1354 in credits a month, and its $200 subscribers to burn $2708 in credits a month. Shellac also adds that Anthropic doesn’t even charge for cache reads, which are charged at around 10% of the cost of tokens. In simpler terms, a $20-a-month subscriber can spend 8.1x their value, and both $100 and $200-a-month subscribers can spend 13.5x. This is very important, because it’s core to Anthropic’s primary business model: deception. It cannot afford to support Claude at this scale, which is why it constantly needs to raise billions of dollars. And when it needs to raise those dollars, Anthropic opens up the floodgates with eased rate limits, paid influencer marketing campaigns, press pushes and, of course, Dario Amodei saying nonsense like that we’re “near the end of the exponential,” and if you’re wondering what that means, that makes two of us. Some genius will claim that “inference is profitable” and that “this is the gym membership model,” and I must be clear how wrong you are. There is no actual proof that inference is profitable — even if it were, it would have to be so profitable that Anthropic can afford to have users spend 500%+ in API calls every single month. It’s actually far simpler. What Anthropic is doing is creating the illusion of a product that can be sold at $20, $100 or $200 a month, when the underlying economics are somewhere in the region of spending anywhere from $8 to $13 to make $1. Anthropic isn’t a business — it’s a parasite that lives off of venture capital and hype. Comment deleted
I'm interested on the open source models that train with claude, is that distillation about? is he talking about the chinese open models? Comment deleted