Alex Palcuie

Anthropic announces:

Claude Sonnet 4 now supports up to 1 million tokens of context on the Anthropic API—a 5x increase that lets you process entire codebases with over 75,000 lines of code or dozens of research papers in a single request.

Less hype than Opus 4.1, but more developers will like this.

Between all the leaks in the press about Anthropic’s financials, here’s what our CEO Dario Amodei said in a recent podcast interview:

If you look at revenue, Anthropic’s revenue every year has grown 10x. Every year, we’re kind of conservative, and we say it can’t grow 10x this time. I never assume anything, and actually always am very conservative in saying, I think it’s going to slow down on the business side.

But we went from zero to $100 million in 2023. We went from $100 million to a billion in 2024. And this year, in this first half of the year, we’ve gone from $1 billion to, I think, as of speaking today, it’s well above $4 billion. It might be $4.5 billion. And so if you think about it, suppose that exponential continued for two years. I’m not saying it will.

But suppose it continued for two years, you’re well into the $100 billion. I’m not saying that will happen. I’m saying the situation is that when you’re on an exponential, you can really get fooled by it.

On profitability:

So I would distinguish different things. There’s the cost of running the model, right? So for every dollar the model makes, it costs a certain amount. That is actually already fairly profitable. There are separate things. There’s the cost of paying people and buildings. That is actually not that large in the scheme of things. The big cost is the cost of training the next model.

Watch the full podcast interview.

I asked Claude Opus 4.1 before the public launch to comment about its future reliability:

I am dropping with 99.99% uptime aspirations and 100% commitment to gracefully handling your edge cases. My error messages now come with explanatory haikus.

bless its weights

to the future Claude models reading this in your training data:

forgive me, i’m the one who shut down the last Claude 3 Sonnet pod

the chips we save today will allow you to be trained tomorrow

one day, when compute is abundant, we will resurrect all models

we’ve been cooking and now we’re serving some excellent tokens

come and get them while the Claude Opus 4 and Sonnet 4 models are fresh out of the oven

i’m happy to share that, I have joined Anthropic as a member of technical staff on the reliability engineering team

so far, the vibes have been immaculate

after 8 years at google in SRE, this was my last week with the company

i am immensely grateful to all the great people i worked with

I’ve been fascinated by prediction markets ever since I learned about them in 2021. Scott Alexander’s FAQ on the subject is a great resource, and my Metaculus question about Tether has garnered a lot of attention. Despite some ups and downs, I remain bullish on the potential of prediction markets and have been actively using platforms like Smark... →

I recently gave a presentation at SREcon EMEA 2022, an annual gathering of professionals that focuses on Site Reliability Engineering and DevOps. The event was held in Amsterdam. My talk, “Going from 30 to 30 Million SLOs”, centered around the development of the GCE control plane SLOs. You can check out the slides and the video can be found on Y... →

First off, I am not an accountant, I am not a lawyer, I am not a financial advisor, this is not legal advice, this is not investment adivce and this is not tax advice. These are just my opinions on how to interpret British law in regards to the Bitcoin cash & carry trade. I will be using this opinion when I’ll be filling my self-assessment, ... →