Azeem Azhar

Let's start with the biggest ledger entry of the day, Anthropic.

Uh, on October fifteenth, Azeem Azhar's Exponential View, uh, published an analysis titled Anthropic: The Agony and the Ecstasy based on leaked numbers from Anthropic's private investor briefing.

The numbers are almost absurdly strong, and yet markets treated them like a disappointment.

According to Azhar's synthesis, Anthropic posted third-quarter revenue of sixteen point eight billion dollars, nearly four times the entirety of twenty twenty-five.

[upbeat music] Hi, I'm Nicholas Thompson, CEO of The Atlantic, and this is The Most Interesting Thing in AI.

A few weeks ago, I had Azeem Azhar on the show, and we talked about one of the biggest questions in AI.

Are we in a bubble? The technology is amazing, but that does not mean the economics work.

And if you've listened to that episode, you know Azeem's answer was, "No, no, it's not a bubble." We worked through a whole series of arguments.

Thank you.

That was Ray Dalio.

If you enjoyed this episode, you might also like my conversation with Azeem Azhar just three weeks ago or my episode with Stanford's Erik Brynjolfsson from May of this year.

And if you want more of this podcast, please subscribe wherever you watch or listen.

Now contrast all of this with biotech, where loops are older, slower, and more disciplined.

The strongest non-AI source today is Azeem Azhar's Exponential View, published October 12th, titled What Is Bigger Than AI? His answer: GLP-1s.

The data is striking.

Exponential View says that in 2024, about three percent of US adults took a GLP-1 for weight loss.

So today's thesis is the end of screen time is not the end of work, it is the beginning of supervision time.

The cleanest signal came this morning from Azeem Azhar's Exponential View, dated October tenth, in an essay titled The End of Screen Time.

Azeem describes arriving in London after forgetting his laptop in Amsterdam.

A year ago, that would have been catastrophic.

That may sound abstract, so let's start with the most mind-bending development of the day.

On October 7th, Azeem Azhar's Exponential View published What is Left to Do? The piece discusses OpenAI releasing a range of mathematical results produced by an unreleased frontier model, seven hundred and twenty-two manuscripts in three hundred and seventy-two families spanning number theory, complexity theory, and mathematical physics.

Azeem says the average result took the equivalent of about three hours of ChatGPT Pro thinking.

He also cites scientist Derya Unutmaz arguing that the body of work represents eighty-one percent of the major math discoveries in the past three years, a claim worth treating carefully.

Which brings us to one of today’s most important stories: legal AI.

On October fifth, Azeem Azhar’s Exponential View published Monday Data: More AI, More Justice.

It’s one of the clearest examples of AI removing friction and immediately creating a new bottleneck.

Azeem notes that in England and Wales, sixty percent of defendants in defended county court claims this year did not have a lawyer.

They wake up worried about systems that advise investors, negotiate contracts, grade students, screen patients, influence children, write code, run cyber operations, and act through accounts.

Azeem Azhar's Exponential View newsletter on October third titled Quick Weekend Reads: The Big AI Questions adds a deeper cultural layer.

Azhar highlights the economist critique of effective altruism as this century's biggest idea, especially the concern that a small group of Silicon Valley aligned thinkers may believe only they can be trusted with powerful AI.

He also points to reporting that Anthropic consulted religious scholars and theologians on machine consciousness, including the Vatican's position that AI systems do not possess bodies, joy, pain, experience, or moral conscience.

The prospectus includes a warning that their products pose an existential risk to humanity.

Azeem Azhar goes over the leaked details here.

They have about five hundred and eighteen billion dollars in compute commitments over seven to ten years, which means they still have work to do on that.

Image.

Ezra Klein's interview of Jensen Huang, which you call an interview for the ages.

And then I thought an equally interesting interview between two people who have appeared on my show before, Nick Thompson, the CEO of Atlantic, and Azeem Azhar, the Golders Green-based technologist, the founder and CEO of The Exponential View.

Let's deal with this Juan Klein interview.

I have to say, I thought it was interesting, but I disagree that...

We will talk next week.

Not a great week for OpenAI or Anthropic, certainly on the marketing point of view.

And I strongly suggest everyone watches Klein's interview with Huang and also Thompson's interview with Azeem Azhar.

They're both excellent and very educational in making sense of this complicated and very fast-moving economy.

So I wanted to get a guest who I could talk to about all of this, kind of run through every argument about whether there's a bubble.

So I called my friend Azeem Azhar.

So Azeem, who runs Exponential View, which is a substack, it's a podcast, he's out there a lot, and he's done a really in-depth report about the economics of AI.

And it is his firm belief that we are not in a bubble.