What do you think?
Rate this book
The Scaling Curve: Dario Amodei, Anthropic, and the Race to Build and Survive Superintelligence
He helped discover the laws that predict how intelligent AI systems will become. Then he realized no one was preparing for what would happen when they did.
In 2021, Dario Amodei walked away from his position as VP of Research at OpenAI—one of the most powerful roles in artificial intelligence—because he believed the organization he helped build was not taking the dangers of its own technology seriously enough. Within weeks, he and his sister Daniela, along with five colleagues, founded Anthropic on a premise that most of Silicon Valley considered that the people most terrified of AI should be the ones building it.
This is the definitive account of that gamble—and of the scientist at its center.
Drawing on extensive research and Amodei's own public statements, Claude St. John traces the arc from a math-obsessed childhood in San Francisco's Mission District to the founding of what Dario has called the fastest-growing software company in history at its scale. Along the way, the book chronicles the discovery of the scaling laws that rewrote the rules of AI, the dramatic departure from OpenAI that sent shockwaves through the industry, the near-catastrophic entanglement with Sam Bankman-Fried's FTX, and the creation of Claude—an AI system trained not just to be capable, but to be good.
At the heart of the story is an extraordinary contradiction. Amodei has warned Congress, world leaders, and the public that AI may be the most dangerous technology humanity has ever created. He has described scenarios in which superintelligent systems could escape human control, destabilize nations, or end civilization. And then, every morning, he goes back to his office and builds it anyway—raising tens of billions of dollars to make his models smarter, faster, more powerful. His defense is not that the danger is overstated. It is that the danger is precisely the reason someone with his understanding must be the one doing it.
Dario Amodei is a story about a brother and sister who bet everything on an idea the world wasn't ready for. It is about what happens when the scaling curves that one physicist discovered in a Baidu research lab begin to reshape the global economy, the balance of power between nations, and the very question of what it means to be intelligent. It is about the handful of people who saw the future coming before anyone else—and who now must decide whether humanity survives its own greatest invention.
Part biography, part intellectual thriller, part warning, this book asks the question that Dario Amodei borrowed from Carl Sagan's Contact and has carried with him for twenty How did you do it? How did you survive this technological adolescence without destroying yourselves?
We don't have the answer yet. This is the story of the man who is trying to find it.
In 2021, Dario Amodei walked away from his position as VP of Research at OpenAI—one of the most powerful roles in artificial intelligence—because he believed the organization he helped build was not taking the dangers of its own technology seriously enough. Within weeks, he and his sister Daniela, along with five colleagues, founded Anthropic on a premise that most of Silicon Valley considered that the people most terrified of AI should be the ones building it.
This is the definitive account of that gamble—and of the scientist at its center.
Drawing on extensive research and Amodei's own public statements, Claude St. John traces the arc from a math-obsessed childhood in San Francisco's Mission District to the founding of what Dario has called the fastest-growing software company in history at its scale. Along the way, the book chronicles the discovery of the scaling laws that rewrote the rules of AI, the dramatic departure from OpenAI that sent shockwaves through the industry, the near-catastrophic entanglement with Sam Bankman-Fried's FTX, and the creation of Claude—an AI system trained not just to be capable, but to be good.
At the heart of the story is an extraordinary contradiction. Amodei has warned Congress, world leaders, and the public that AI may be the most dangerous technology humanity has ever created. He has described scenarios in which superintelligent systems could escape human control, destabilize nations, or end civilization. And then, every morning, he goes back to his office and builds it anyway—raising tens of billions of dollars to make his models smarter, faster, more powerful. His defense is not that the danger is overstated. It is that the danger is precisely the reason someone with his understanding must be the one doing it.
Dario Amodei is a story about a brother and sister who bet everything on an idea the world wasn't ready for. It is about what happens when the scaling curves that one physicist discovered in a Baidu research lab begin to reshape the global economy, the balance of power between nations, and the very question of what it means to be intelligent. It is about the handful of people who saw the future coming before anyone else—and who now must decide whether humanity survives its own greatest invention.
Part biography, part intellectual thriller, part warning, this book asks the question that Dario Amodei borrowed from Carl Sagan's Contact and has carried with him for twenty How did you do it? How did you survive this technological adolescence without destroying yourselves?
We don't have the answer yet. This is the story of the man who is trying to find it.
255 pages, Kindle Edition
Published February 21, 2026
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 13 of 13 reviews
April 5, 2026
Basic account of CEO and co-founder of Anthropic. Regardless of lack of new information, any content on one of the most important companies in the world, as AI inflects, is invaluable.
March 16, 2026
interesting read
Follows Dario as he leaves open AI to found Anthropic. Blends tech insights with high stakes narrative - is AI our future? Are we ready for it?
Follows Dario as he leaves open AI to found Anthropic. Blends tech insights with high stakes narrative - is AI our future? Are we ready for it?
May 10, 2026
This book cropped up recently on Amazon's best-seller list for AI books. Only after snagging a copy and reading twenty or so pages of what looked like fine journalism did I find out, to my shock, that this 350-page paperback had apparently been belched out via AI over the course of two weekends. Egads! I decided to read it anyway.
The book, about Anthropic CEO Dario Modei, was the brainstorm of Anthropic employee who used dozens of papers, speeches, and interviews of Modei's as data feedstock. The story was about how Modei had as a young employee of Baidu and OpenAI had been one of the first to discover that that the striking improvements in LLM's answers over its early versions seemed to correlate well with the growth in the amount of trained data, the number of "parameters" (numeric threshold values used in demarcating patterns), and computing speed / bandwidth. Modei was the first to formalize and promote the idea that LLMs would just improve as these three hyperparameters grew, with little apparent need for customary extra levels of buggy engineering. The surprising simplicity of the LLM architecture, iteratively predicting the next word in an answer, combined with multi-level neural nets, numerical data to register human feedback during training, and specialized Nvidia GPU microchips, could apparently sidestep the kinds of problems and failures of the 1970's and 90's that had left the overly fragile AI of those times in prolonged deep freezes.
It's unclear how much or what kinds of help the Anthropic employee applied during those two weekends, but the result, a cohesive, entertaining, high-octane narrative, is quite remarkable. The result is a smooth, pleasant read, economical yet full of well-organized information, the style silky yet sinewy (or maybe the other way around). It's well worth reading if only to calibrate what AI as author can do now. (For me that experience was, no surprise, both exhilarating and sobering.)
As for the scaling curve itself, juicing it to new heights has become the norm (new data centers coming soon to a town near you), but rumor has it doubts are being raised about diminishing returns: scaling curve may not have the legs of Moore's Law. This might be a good thing since the current exponential growth and cost rates are unsustainable. On the other hand, financial markets may be at serious risk if and when tech giants find themselves unable to service debts now in the trillions. It could be a bumpy ride. But as always these days, we do live in such interesting times.
The book, about Anthropic CEO Dario Modei, was the brainstorm of Anthropic employee who used dozens of papers, speeches, and interviews of Modei's as data feedstock. The story was about how Modei had as a young employee of Baidu and OpenAI had been one of the first to discover that that the striking improvements in LLM's answers over its early versions seemed to correlate well with the growth in the amount of trained data, the number of "parameters" (numeric threshold values used in demarcating patterns), and computing speed / bandwidth. Modei was the first to formalize and promote the idea that LLMs would just improve as these three hyperparameters grew, with little apparent need for customary extra levels of buggy engineering. The surprising simplicity of the LLM architecture, iteratively predicting the next word in an answer, combined with multi-level neural nets, numerical data to register human feedback during training, and specialized Nvidia GPU microchips, could apparently sidestep the kinds of problems and failures of the 1970's and 90's that had left the overly fragile AI of those times in prolonged deep freezes.
It's unclear how much or what kinds of help the Anthropic employee applied during those two weekends, but the result, a cohesive, entertaining, high-octane narrative, is quite remarkable. The result is a smooth, pleasant read, economical yet full of well-organized information, the style silky yet sinewy (or maybe the other way around). It's well worth reading if only to calibrate what AI as author can do now. (For me that experience was, no surprise, both exhilarating and sobering.)
As for the scaling curve itself, juicing it to new heights has become the norm (new data centers coming soon to a town near you), but rumor has it doubts are being raised about diminishing returns: scaling curve may not have the legs of Moore's Law. This might be a good thing since the current exponential growth and cost rates are unsustainable. On the other hand, financial markets may be at serious risk if and when tech giants find themselves unable to service debts now in the trillions. It could be a bumpy ride. But as always these days, we do live in such interesting times.
April 20, 2026
I picked this book to read after reading The Infinity Machine:
Demis Hassabis, DeepMind, and the Quest for Superintelligence, by Sebastian Mallaby. The Infinity Machine was more deeply reported and covered the development of AI from a much earlier stage. Both books are highly favorable to the person at the book’s center; either Demis Hassabis or Dario Amodei. I wish both books had a more objective viewpoint of the personalities they covered.
The Infinity Machine book covered the path that led from the DeepMind company in England to Google Gemini. That history started earlier, covered the early work that led to AI, and had a more complete biography of Demis Hassabis. This book has a less complete biography of Dario Amodei and focuses on the development of Anthropic and Claude. However, the book does cover the more mature development of AI models and goes into much more detail on the efforts to make AI models safer. The AI explanations are presented in easy-to-understand analogies. The book is fairly short and easy to read. I give it high marks and would recommend it to anyone interested in AI.
Demis Hassabis, DeepMind, and the Quest for Superintelligence, by Sebastian Mallaby. The Infinity Machine was more deeply reported and covered the development of AI from a much earlier stage. Both books are highly favorable to the person at the book’s center; either Demis Hassabis or Dario Amodei. I wish both books had a more objective viewpoint of the personalities they covered.
The Infinity Machine book covered the path that led from the DeepMind company in England to Google Gemini. That history started earlier, covered the early work that led to AI, and had a more complete biography of Demis Hassabis. This book has a less complete biography of Dario Amodei and focuses on the development of Anthropic and Claude. However, the book does cover the more mature development of AI models and goes into much more detail on the efforts to make AI models safer. The AI explanations are presented in easy-to-understand analogies. The book is fairly short and easy to read. I give it high marks and would recommend it to anyone interested in AI.
May 20, 2026
Interesting person, fascinating topic — that’s what drew me to this book.
The first half captured my attention immediately; I read it within a few hours. In particular, the background stories about the founders of Anthropic offered several interesting insights I hadn’t seen before.
However, as the book progressed, it increasingly felt like a compilation of Dario Amodei’s essays and podcast appearances rather than an original piece of work. Many of the key ideas were repeated multiple times, which made the second half considerably less engaging.
What I found especially disappointing was the apparent lack of genuine original research. There seem to be no exclusive interviews with Dario, Daniela, or other major protagonists involved in the story. At times, I even wondered whether parts of the book were AI-generated — or at least heavily AI-assisted. If so, the editorial process clearly did not do the material justice.
That is unfortunate, because the subject itself is highly relevant and state-of-the-art. A topic of this importance deserved a much deeper and more thoughtfully crafted treatment.
The first half captured my attention immediately; I read it within a few hours. In particular, the background stories about the founders of Anthropic offered several interesting insights I hadn’t seen before.
However, as the book progressed, it increasingly felt like a compilation of Dario Amodei’s essays and podcast appearances rather than an original piece of work. Many of the key ideas were repeated multiple times, which made the second half considerably less engaging.
What I found especially disappointing was the apparent lack of genuine original research. There seem to be no exclusive interviews with Dario, Daniela, or other major protagonists involved in the story. At times, I even wondered whether parts of the book were AI-generated — or at least heavily AI-assisted. If so, the editorial process clearly did not do the material justice.
That is unfortunate, because the subject itself is highly relevant and state-of-the-art. A topic of this importance deserved a much deeper and more thoughtfully crafted treatment.
April 4, 2026
First things first: This book is written by AI. There was extensive fact checking and editorial prompts given to the AI model. In my opinion you recognize that it was not written by a human by the prose. It just feels a little off.
The book itself is interesting and I like the premise that AI should be programmed by the people that are most scared of it.
I did expect to read more about Dario Amodei than what was actually discussed, but I will have a look at his published essays.
I would have wished for a more clear written by AI label and maybe a foreword where the writing process was described.
I recived an ARC of this book trough NetGalley in exchange for an honest review.
The book itself is interesting and I like the premise that AI should be programmed by the people that are most scared of it.
I did expect to read more about Dario Amodei than what was actually discussed, but I will have a look at his published essays.
I would have wished for a more clear written by AI label and maybe a foreword where the writing process was described.
I recived an ARC of this book trough NetGalley in exchange for an honest review.
May 12, 2026
Essential Reading!
The best book in print to begin to comprehend the promise and peril of artificial intelligence (AI). Learn why the founders of Anthropic are the most thoughtful and moral scientists and leaders to trust concerning the policies and governance that is necessary to guide the future of this world changing technology. Every legislator and every informed citizen needs to read this book. Highly recommended.
The best book in print to begin to comprehend the promise and peril of artificial intelligence (AI). Learn why the founders of Anthropic are the most thoughtful and moral scientists and leaders to trust concerning the policies and governance that is necessary to guide the future of this world changing technology. Every legislator and every informed citizen needs to read this book. Highly recommended.
May 24, 2026
Still about a quarter of the way in, but this "AI slop" is a good read that I am learning a lot from, and mind you, I have a PhD in computer engineering. And it took me a bit to realize that it was written with Claude LLM, one of the best chatbots out there
Make sure you read it critically though. Many statements in the book are open to debate. Some are, to my mind, mostly wrong. But overall, it's informative about Dario Amodei's background and the last decade of amazing progress in AI.
Make sure you read it critically though. Many statements in the book are open to debate. Some are, to my mind, mostly wrong. But overall, it's informative about Dario Amodei's background and the last decade of amazing progress in AI.
This entire review has been hidden because of spoilers.
May 30, 2026
AI-generated slop. I would not recommend this book.
It’s repetitive, poorly edited, and offers very little original insight. The biggest issue for me is the lack of transparency about the use of AI. The book is marketed and presented as if it were written by a human, which I find misleading.
I have no problem with AI-assisted writing, but I do expect honesty about how a book was created. For that reason alone, I can’t recommend it.
It’s repetitive, poorly edited, and offers very little original insight. The biggest issue for me is the lack of transparency about the use of AI. The book is marketed and presented as if it were written by a human, which I find misleading.
I have no problem with AI-assisted writing, but I do expect honesty about how a book was created. For that reason alone, I can’t recommend it.
May 31, 2026
This book reads like a love story to Dario and Anthropic. The only criticism comes from Dario's essay and then it sounds like he is the only one that worries about security. It's a rather boring read, especially after 40% where the author (or AI?) thought it was a good idea to repeat everything on every other page.
May 11, 2026
It does not seem to provide anything new other than what is out there. Plus it felt AI written or not reviewed carefully as contents are repeated quite frequently, for no obvious reason.
June 6, 2026
3.5 like a long form Atlantic article as a book
May 4, 2026
Well written
This is an interesting one. It is a really well written book.
However, as I was reading it, two things came. To mind:
- The author did not have access to Amodei or interviews for this book
- The author really admires him
I think if the author would have had access to interview and talk with Amodei it would have a 5*
This is an interesting one. It is a really well written book.
However, as I was reading it, two things came. To mind:
- The author did not have access to Amodei or interviews for this book
- The author really admires him
I think if the author would have had access to interview and talk with Amodei it would have a 5*
Displaying 1 - 13 of 13 reviews