What are you talking about?

The Attention is all you need paper has Google logo, not character.ai

It's not more expensive at all. They are all open weights models. I run them on 2x8xH100. They cost the same.

Openrouter has them as significantly more expensive.

Source? Their website says 1000t/s https://www.cerebras.ai/blog/which-is-faster-gemini-3-5-flas...

This is likely correct, sorry for the bad info. Was working from memory.

The deal ends in June 2029, which is before 2030.

They also might switch power sources.

So you mean to say Google knows AI definitely won't be needed after June 2029... Understood...

Indexes have rules for entry, so people using them supposedly agree with them. On the other hand, they were talking about opening exceptions for these companies, which was not part of the initial rules.

That's not even slightly what they said. They said that people can trade the stock directly if they want to, rather than have it essentially forced upon them by inclusion in an index

When I grew up in Singapore, the locations of all data centers are secret by default to prevent terrorist attacks.

I mean the framing is so silly.

"If bank vaults are so great, why would you not advertise their locations". It's a mystery, is what it is.

Well, banks aren't kept secret and the vaults are usually inside.

The point is that there is zero positive correlation between a vault doing its job and telling everyone where exactly it is. The bank would happily have it hidden in different secret dimension, if they could. It makes the vault safer. It's better.

"Why would you not tell us how to get to the secret dimension if it's so great?" It's just a dumb question.

> The point is that there is zero positive correlation between a vault doing its job and telling everyone where exactly it is.

Yes, but the more important correlation is between keeping something secret and it being a bad idea (at least for some people). It's like using security through obscurity - it has to be kept secret because it's flawed, whereas a bank vault doesn't need to be in a secret location because it has other means of protecting its contents.

There's so many examples of closed source "security" that is just a clusterfuck. Not that mistakes aren't made with open source, but knowing that other people will likely look at your code tends to make people put a bit more thought into it.

That confuses how offense/defense works in software vs real world. Publishing your vault layout will not lead to a lot of people want to use the vault and in the process submitting "patches" that you then can use and make your vault better. A built vault is not something that is easily malleable.

More people can however work on a great plan to attack it, which only has to work once, to be worth it.

If every employee is part of a union, what happens then when companies over-hire?

Unionized companies can still do layoffs.

How is SpaceX not making money?

Total investment is 20-40B, rent to Anthropic for 45B over 3 years.

Anthropic is also profitable now.

Anthropic is not profitable and does not expect to be revenue even until 2028. They are just losing less billions per year than everyone else.

As always, free cash flow and unit economics are more interesting than net profit.

By which metrics Anthropic is making a lot of money.

how exactly can you assume they have unit profitability

Plenty of analysis on this point, the inputs are not secret. Check out Semianalysis for example.

They have something like 70% margin on inference.

They had 40% margin on inference in 2025.

https://archive.is/aKFYZ

You can quibble about the exact numbers, but I think it’s fairly clear at this point that inference is profitable with decent margins. Like you say, unit economics are more interesting than the profitability of the company as a whole.

so speculation from a semiconductor website. that's incredibly likely to be wrong

I think you should do a bit of research and come back once you have fixed your knowledge gaps.

what research? it's private and there is large incentive to lie until ipo

We will see once Anthropic releases their S-1 too this summer.

Closer to 18b than 40. Running costs are 1-2b a year.

More like $25 billion since 2025, with $7 billion of spending in the past 3 months. Look at page 22 of the filing.

Why didn't other CEOs buy them to wreck Zuck's platform?