Is this an earnest reply or a Bluesky-missing-the-joke thing?

Is this a Nazi thing or just a German cuisine thing?

I’ll ask for this 🤏 much more wasabi and see if that elicits a reaction to confirm, brb

I haven’t done a full phenological analysis to confirm but I strongly suspect he is Korean (slightly pudgy but more elongated than rounded face, which rules out Chinese, and his accent is bad enough I doubt he’s Japanese)

Imagine looking at an index with basis rolldown included, frankly you're the pervert here

“This is great, I get to bend down for a cool drink of water and reach up to pick fresh fruit whenever I want it”

That's not normal rich person behavior already? How do they get their points?

No, CC charge offs are like high single digits. The reason interest rates are so high is because that means that they have essentially first priority because consumers will pay off 30% balances first so it’s kind of a first lien.

There’s already an existing biz model where you take all the credit risk out up front via interchange fees, it’s called BNPL and they charge like 8% to the merchant lol

American consumers, the greatest spenders on God’s green earth, don’t spend based on meaningless things like “income” or “net worth”, but rather on current cashflow and credit, and so any cuts to limits will flow through very quickly to spending.

A trillion dollars of actual spending isn’t that much money, they can probably digest that pretty easy in this scenario. If the fake multi-trillion spending turns out to be real instead of just press releases then they could be in trouble.

The premium pricing isn’t for the AI model, it’s for the human sales rep to take you to steak dinners on the corporate card, same as it ever was

Unlike XAI which loses the most money to their least valuable customers with the lowest willingness to pay (people generating videos), Anthropic loses the most money to their most valuable customers with the highest willingness to pay (CC superusers). That is a very easy business problem to fix!

To borrow another industry analogy, AI model's future will look like airlines, where a bunch of exquisite pricing algos sort you into dozens of different fare classes assigning precisely how many free checked bags and biscoff cookies you get with your ticket and loyalty membership.

I suspect that's almost entirely due to the growth of test-time compute from reasoning models, which is by definition something that you can dial up/down (or charge on a per unit basis) as necessary. What they need is price discrimination, and businesses are usually quite good at that.

Sam is certainly going to lose a margin war with Google over token pricing but even if he ends up McClendoning himself and OAI the industry won’t “go away”, especially now that models can update their knowledge via cheap search queries instead of expensive training runs so useful life is ~indefinite

Google, Tencent, and Alibaba won’t ever run out of money for training runs, OAI and Anthropic might and have to pivot to inference only on existing models which is much cheaper, you probably get something akin to the 2015 shale bust where token prices fall to cash cost because people keep “pumping”