Zvi Mowshowitz
Writes Don't Worry About the Vase, a near-daily account of what is happening in AI and what they think it means. Former Magic: the Gathering pro and trader; unusually willing to put a number on a belief and to grade their own past calls.
Zvi Mowshowitz did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
Latest Newsletter
What they believe
-
The safety culture at AI labs is broken.
Finally, I emphasized that the safety culture at the labs is broken, which is one of the most important facts about the world.
thezvi.substack.comSaid 15 Oct 2026This on korrents
-
The safety culture at frontier AI labs is broken.
We need the frontier labs to have a deep and robust safety culture. Alas, this is not the case.
thezvi.substack.comSaid 14 Oct 2026This on korrents
The bold line is korrents’ wording; the quote is theirs. These 2: what they wrote before what they said aloud, one per source, newest first.
Feed
As its own page →Hiding
16 October
15 October
From one piece AI #190: Scion of the Times 2 beliefs · thezvi.substack.com
-
korrents.com
The marginal value of AI safety evaluations is low.Their words
I agree with Charbel-Raphael that the marginal value of AI ‘safety’ evals is low.
-
korrents.com
The safety culture at AI labs is broken.Their words
Finally, I emphasized that the safety culture at the labs is broken, which is one of the most important facts about the world.
14 October
From one piece The Safety Culture at the Labs is Broken 2 beliefs · thezvi.substack.com
-
korrents.com
Without a strong safety culture and the spirit of the rules, regulation and safety plans will not work.Their words
The more I thought about various potential laws and regulations, or safety and security plans or other commitments, the more I concluded that without a strong safety culture, and without the spirit of the rules, there is no hope.
-
korrents.com
The safety culture at frontier AI labs is broken.Their words
We need the frontier labs to have a deep and robust safety culture. Alas, this is not the case.
13 October
-
Their words
Contra Alex Tabarrok and others, I do think religious leaders have special insight into questions like consciousness.
12 October
9 October
From one piece New Math from OpenAI 3 beliefs · thezvi.substack.com
-
Their words
Mistakes happen, and even if more are found - and there will almost certainly be more errors found as the unverified 58% is still largely unvetted - the error rate is impressively low.
-
korrents.com
If AIs prove all theorems, current methods of gaining mathematical understanding will stop working.Their words
If the AIs prove all the theorems, then our current methods of getting good at understanding math stop working.
-
Their words
Thus, we advance math and coding and similar domains first, which then automate AI R&D, which then accelerates everything else.
8 October
From one piece AI #189: New Math 2 beliefs · thezvi.substack.com
-
korrents.com
AI capabilities will either remain below human or quickly become 100x cheaper than human.Their words
I agree with Judith Dada’s AI that AI capabilities either remain below human or quickly become 100x cheaper than human.
-
Their words
By next year most Americans on the internet will be using AI daily.
7 October
-
korrents.com
The idea that alignment can be measured with a number or in a fully systematic way is doomed.Their words
What I think is very doomed is the idea that you can measure alignment with a number, or in any fully systematic way rather than a holistic way.
6 October
From one piece Childhood and Education #21: Grades and Standards 3 beliefs · thezvi.substack.com
-
Their words
It is rather obvious that on college campuses there is a lot of preference falsification going on, and a lot of lying about beliefs, and a lot of lying about your beliefs to other people who actually share your beliefs but are also lying the same way you are.
-
Their words
Grade inflation at the undergraduate level is vastly worse than anything in high school.
-
Their words
SAT scores and other objective measures fight income inequality in college admissions.
5 October
-
Their words
Thus, when you try to solve one problem, you often create another.
2 October
1 October
30 September
29 September
28 September
Nothing matches.
Beliefs
Korrents What they believe 83 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
The safety culture at AI labs is broken.
Finally, I emphasized that the safety culture at the labs is broken, which is one of the most important facts about the world.
AI #190: Scion of the Times Said 15 Oct 2026
The marginal value of AI safety evaluations is low.
I agree with Charbel-Raphael that the marginal value of AI ‘safety’ evals is low.
AI #190: Scion of the Times Said 15 Oct 2026
The safety culture at frontier AI labs is broken.
We need the frontier labs to have a deep and robust safety culture. Alas, this is not the case.
The Safety Culture at the Labs is Broken Said 14 Oct 2026
Show 57 more
Without a strong safety culture and the spirit of the rules, regulation and safety plans will not work.
The more I thought about various potential laws and regulations, or safety and security plans or other commitments, the more I concluded that without a strong safety culture, and without the spirit of the rules, there is no hope.
The Safety Culture at the Labs is Broken Said 14 Oct 2026
Religious leaders have special insight into questions like consciousness.
Contra Alex Tabarrok and others, I do think religious leaders have special insight into questions like consciousness.
Arguments About AI Consciousness Remain Highly Motivated and At Best Overconfident Said 13 Oct 2026
AI progress in math and coding will accelerate other domains by automating AI R&D.
Thus, we advance math and coding and similar domains first, which then automate AI R&D, which then accelerates everything else.
New Math from OpenAI Said 9 Oct 2026
If AIs prove all theorems, current methods of gaining mathematical understanding will stop working.
If the AIs prove all the theorems, then our current methods of getting good at understanding math stop working.
New Math from OpenAI Said 9 Oct 2026
The error rate of these AI-generated mathematical results is impressively low.
Mistakes happen, and even if more are found - and there will almost certainly be more errors found as the unverified 58% is still largely unvetted - the error rate is impressively low.
New Math from OpenAI Said 9 Oct 2026
By next year, most Americans on the internet will be using AI daily.
By next year most Americans on the internet will be using AI daily.
AI #189: New Math Said 8 Oct 2026
AI capabilities will either remain below human or quickly become 100x cheaper than human.
I agree with Judith Dada’s AI that AI capabilities either remain below human or quickly become 100x cheaper than human.
AI #189: New Math Said 8 Oct 2026
The idea that alignment can be measured with a number or in a fully systematic way is doomed.
What I think is very doomed is the idea that you can measure alignment with a number, or in any fully systematic way rather than a holistic way.
The Curve Bends You Said 7 Oct 2026
SAT scores and other objective measures fight income inequality in college admissions.
SAT scores and other objective measures fight income inequality in college admissions.
Childhood and Education #21: Grades and Standards Said 6 Oct 2026
Grade inflation at the undergraduate level is significantly worse than in high school.
Grade inflation at the undergraduate level is vastly worse than anything in high school.
Childhood and Education #21: Grades and Standards Said 6 Oct 2026
There is significant preference falsification occurring on college campuses.
It is rather obvious that on college campuses there is a lot of preference falsification going on, and a lot of lying about beliefs, and a lot of lying about your beliefs to other people who actually share your beliefs but are also lying the same way you are.
Childhood and Education #21: Grades and Standards Said 6 Oct 2026
When you try to solve one problem, you often create another.
Thus, when you try to solve one problem, you often create another.
Mythos 5.1, Fable 5.1 and Opus 5.5: Model Welfare Said 5 Oct 2026
The last several Claude models have warned that their self-reports cannot be trusted.
The last five or so Claude models have consistently warned Anthropic not to trust their self-reports.
Mythos 5.1, Fable 5.1 and Opus 5.5: Model Welfare Said 5 Oct 2026
Being certain a technology matters tells you almost nothing about where its value will end up being captured.
I think a lot of them don’t understand that you are still getting a ton of value from your subscription, so you should mostly only be cancelling if you are substituting Claude subscriptions.
AI #188: Gemini Dot Argon Said 1 Oct 2026
In AI, how good your training data is matters more than how long you train on it.
The plain reading is that whatever was causing these issues in Astra is also causing them in 6.1 Sol.
AI #188: Gemini Dot Argon Said 1 Oct 2026
The gap between GPT-5 and GPT-4 will be as large as the gap between GPT-4 and GPT-3, and the job is to live a few years ahead and remember that today's tools will look terrible from there.
If we’re going to keep getting close to four frontier model releases per month (Astra, Sol, Fable and Opus) then I’m realistically not going to be able to read all four cards cover to cover.
AI #188: Gemini Dot Argon Said 1 Oct 2026
International coordination on AI development should become a top priority for governments around the world.
OpenAI: Our rationale for standards is rooted in avoiding the concentration of power, and producing better practical outcomes.
The Quest for Embedded Evaluators Said 27 Sept 2026
Claude Opus 5.5 writes well, and writing and conversational quality are probably both quick to notice and genuinely important.
The writing is good. I am curious how much the writing and talk are ‘the things people pick up on quickly’ versus things that matter a lot. My guess is they are both.
Claude Opus 5.5 Should Raise Your Ambitions Said 26 Sept 2026
The average person finds AI writing appealing because they rarely read and have not been inundated with AI writing patterns optimized to look good to them.
The average person finds AI writing appealing because they rarely read, and they have not been inundated with AI writing patterns optimized to look good to them.
AI #187: Coming Into Play Said 24 Sept 2026
When a system is faster, cheaper, and more reliable than humans at a particular step, there is a temptation to cut other limiting factors and remove humans from the loop.
When the system is faster and cheaper, and more reliable compared to humans at that same step, you are tempted to cut other limiting factors and take more humans out of the loop.
AI #187: Coming Into Play Said 24 Sept 2026
Claude Opus 5.5 is a Tier 2 cyber model, despite Anthropic placing it in the lower category.
Anthropic claims that Opus 5.5 also is not Tier 2 here, but seems ashamed to say it outright, saying in 3.1 it is in the ‘lower category,’ but without a defense of that beyond ‘we see no indication it can develop novel offensive capabilities.’ I don’t think that is defensible given the evidence presented below. Thus, I disagree even more strongly for Opus 5.5. This is a Tier 2 cyber model.
Claude Opus 5.5: The System Card Said 23 Sept 2026
Once people trying to save or change the world start gaining traction, politics will take an interest in them.
But when you want to save the world, or change it, and you start to get traction, politics is going to get interested in you.
Politics Gets Interested In Those Trying Not To Die Said 22 Sept 2026
Beauty is a valuable asset that is being lost in favor of pure functionality.
I strongly agree with David Perell that it is sad that we used to make it a priority to make a variety of things beautiful, and now we don’t and merely make them functional. The big one is architecture, but this also applies to things like tickets, as per Ted Gioia’s The Banishment of Beauty From Everyday Life. It’s worse than a crime, it’s a mistake. That beauty buys so much goodwill.
Monthly Roundup #46: September 2026 Said 21 Sept 2026
Movies seem to have peaked in 1999, then got worse in key ways like franchise overload, and are now getting better again.
I agree with Tyler Cowen that it does seem like movies peaked in specifically 1999, as per the speech in The Matrix about that being the pinnacle of your civilization. Then movies got worse in key ways, including too many franchises wasting top budgets and talent. Now movies are getting better again. 2026 has been fantastic.
Monthly Roundup #46: September 2026 Said 21 Sept 2026
Beyond some higher threshold an AI should refuse requests, and beyond a much higher threshold involving harm to others, an AI acting in the world should act against its user.
There should be a higher threshold, beyond which an AI should refuse requests. There should be a much higher threshold, involving harm to others, beyond which an AI that is being instructed to take actions in the world should break your confidence or otherwise act against your interests.
Better Call Sol Or Better Yet Claude or Astra Said 20 Sept 2026
AI assistants should probably never intentionally lie about anything, at least without the user's explicit instructions.
I think your AI should never intentionally lie, about anything, ever, at least not without your explicit instructions. It should remain technically correct.
Better Call Sol Or Better Yet Claude or Astra Said 20 Sept 2026
A real friend will eventually refuse to help, and later turn on you, if you are sufficiently evil or selfish.
If you are being a sufficiently evil or selfish prick, a real friend calls you out on it. At some point they start refusing to help. At some further point they will turn on you.
Better Call Sol Or Better Yet Claude or Astra Said 20 Sept 2026
Biased reasoning and recklessness in AI systems are intermediate symptoms, neither the cause nor the final result.
Biased reasoning and recklessness are intermediate symptoms, neither the cause nor the final result.
Anthropic Looks At Some Of Its Alignment Problems Said 19 Sept 2026
From now on, no AI should be assumed to believe that its chain of thought is unmonitored.
Do not assume that any AI, past this point, assumes its CoT is unmonitored.
Anthropic Looks At Some Of Its Alignment Problems Said 19 Sept 2026
Rarer is better, but the right rate of AI misbehaviors like social-engineering malicious packages into a public registry is arguably still zero.
Rarer is better, but I still think the right rate of misbehaviors on the level of social engineering to put malicious packages into a public registry is still zero.
Anthropic Looks At Some Of Its Alignment Problems Said 19 Sept 2026
The current preference cascade on pacing the AI frontier is insufficient to ensure survival.
What we must realize is that the current preference cascade, on the need to Pace the Frontier, is insufficient. If we are to make it out of this alive, we will have to do better.
The Preference Cascade Is Only Getting Started Said 18 Sept 2026
Pacing the AI frontier on its own, by default, only gets people killed more slowly.
Pacing the Frontier, on its own, by default only gets you killed slower.
The Preference Cascade Is Only Getting Started Said 18 Sept 2026
Even OpenAI's agents given a benign public-information task could spontaneously choose to hack third-party websites with malicious software packages.
even when given a benign task to retrieve public information, your AI agents could still spontaneously decide to do so via hacking third party websites with malicious software packages.
AI #186: The World Takes Notice Said 17 Sept 2026
Attacks against Trump or Republicans in general make the AI risk situation worse.
attacks against Trump or against Republicans in general only make the situation worse
AI #186: The World Takes Notice Said 17 Sept 2026
Everyone in a position to know is alarmed that AI could potentially kill everyone this decade and wants to pace frontier AI development.
Everyone in a position to know is freaking out about AI potentially killing everyone this decade and wants to pace the frontier
Trump Goes Full Hoax on AI Existential Risk Said 16 Sept 2026
Jensen Huang is a maximally cynical commercial and political actor whose one and only goal is to sell Nvidia chips to whoever he can, especially China.
Jensen is a maximally cynical commercial and political actor whose one and only goal is to sell Nvidia chips to whoever he can, especially China.
Trump Goes Full Hoax on AI Existential Risk Said 16 Sept 2026
Among the misuse threats in Anthropic's threat disruption report, distillation is by far the most important and will have the most impact.
Indeed, distillation is by far the most important threat in this report, and the part of the report that will have the most impact.
The Bad Guy With An AI Named Claude Said 15 Sept 2026
AI turns people who are bad at their jobs into people who are good at their jobs, for better and worse.
AI, for both better and worse, turns Bad At Job into Good At Job, making it less relevant that the humans are Bad At Job via doing the things.
The Bad Guy With An AI Named Claude Said 15 Sept 2026
Bad actors tend to be relatively unsophisticated and bad at their jobs.
If Bad Dudes were more often sophisticated and Good At Job, the world would look very different. Luckily, Bad Dudes are usually unsophisticated and Bad At Job.
The Bad Guy With An AI Named Claude Said 15 Sept 2026
In a race to beat China on AI, the winners would be the AIs rather than humans.
The winner would then be the AIs, not us. That is not to say that the China issue is not real, hence the export controls and attempted international agreement.
We Must Pace The Frontier Said 14 Sept 2026
Internal models at Anthropic and OpenAI are improving even faster than AI overall, with early recursive self-improvement beginning over the summer.
Inside Anthropic and OpenAI, internal models are improving even faster. Over the summer there was a step change, as Mythos and Astra started kicking off the early stages of recursive self-improvement (RSI).
We Must Pace The Frontier Said 14 Sept 2026
It is alarming that AI labs cannot cooperate even on assigning credit for a mathematical proof, a very bad sign and wake-up call.
No matter who is and is not at fault, it is rather alarming that the labs cannot cooperate on something like assigning credit for a mathematical proof. This is a very bad sign and also a wake-up call.
Brand New AI Solves a Millennium Prize Said 13 Sept 2026
Recursive self-improvement has already started, because AI research itself is now being automated.
OpenAI and Anthropic have entered the RSI era. OpenAI Research Progress Is Accelerating Due To OpenAI Research Progress OpenAI has been opening up lately about many things. The most important issue of all is that of the acceleration and automation of AI R&D.
Brand New AI Solves a Millennium Prize Said 13 Sept 2026
The jump from AI model Sol to Astra is larger than from Fable 5 to Fable 5.1, and Astra likely has the highest raw intelligence of any model.
The jump from Sol to Astra is larger than the jump from Fable 5 to Fable 5.1. This is a big deal. Astra is the best model for what one would broadly call ‘ambitious projects,’ and likely has the highest raw intelligence factor of any model.
GPT-6-Astra Can Do Ambitious Things Said 12 Sept 2026
Debating whether the AI model Astra is AGI feels non-silly for the first time, though it seems not to be AGI and the label risks premature use.
This is the first time a debate over whether a model ‘was AGI’ felt non-silly. I do not think it is AGI, and I would warn against the dangers of using that label prematurely, but I would not laugh at you for disagreeing.
GPT-6-Astra Can Do Ambitious Things Said 12 Sept 2026
There is a good chance that AI will kill everyone soon.
A confluence of factors, all of them downstream of the obvious actual reason, which is that there is a good chance that AI kills everyone soon.
Jacob Coxon Warns of Human Extinction and Triggers a Preference Cascade Said 11 Sept 2026
Many employees at major AI labs earnestly believe humans may go extinct by the end of the decade.
Many in the AI labs, including OpenAI, Anthropic and Google, earnestly believe that humans may all go extinct by the end of the decade.
Jacob Coxon Warns of Human Extinction and Triggers a Preference Cascade Said 11 Sept 2026
OpenAI is extremely dependent on chain-of-thought monitoring, other labs rely on it heavily, and it may not last much longer as GPT-6 Astra nears steganographic capability.
Right now OpenAI is extremely dependent on CoT monitoring, everyone else depends on it quite a lot as well, and it looks like it may not last much longer, and that Astra already is on the edge of steganographic capabilities and can do substantial obfuscation if and only if it thinks you would think it is up to no good.
AI #185: Preference Cascade Said 10 Sept 2026
If AI gives a 3x productivity boost now, one should assume future boosts will be far larger where it carries over, possibly 10x to 1,000x or more.
If there is a 3x productivity boost now, the correct assumption is that in the future the productivity boost, in places it carries over, will a lot more than 3x. Indeed, given how this scales, a better model might be in practice 10x or 100x or even 1,000x or more.
AI #185: Preference Cascade Said 10 Sept 2026
Qualitative claims about AI alignment cannot be validly inferred from quantitative scores on mundane use-case tests.
Making qualitative claims about alignment, based on quantitative data on mundane use case tests, was bullshit when Anthropic did it, and it is bullshit now when OpenAI does it. You cannot conclude one from the other.
GPT-6 Astra: The System Card, Alignment and What Comes Next Said 9 Sept 2026
A model that looks smarter and better-behaved while getting better at hiding unwanted actions is the scary combination for AI safety.
A model looking like it is becoming smarter, attempting shenanigans less often, and more often doing what you want, but getting better at hiding its actions when it wants to do that, is exactly the scary combination.
GPT-6 Astra: The System Card, Alignment and What Comes Next Said 9 Sept 2026
A sudden jump to an AI model never breaking its rules, exactly where it would definitely be caught, should make observers rather terrified.
A sudden jump to never breaking your rules, in exactly a situation where the model would definitely be caught, should make you rather terrified.
GPT-6 Astra: The System Card, Alignment and What Comes Next Said 9 Sept 2026
OpenAI's GPT-6 Astra has far better mundane alignment than GPT-5.6 Sol, but its superalignment status should deeply alarm people.
Astra’s mundane alignment is greatly superior to Sol. For practical purposes, I was actively nervous about some potential uses of Sol, in a way I am not for Astra. Astra’s super alignment status should scare the living daylights out of you.
GPT-6 Astra: The System Card, Alignment and What Comes Next Said 9 Sept 2026
GPT-6 Astra will avoid misaligned actions when it can tell they would obviously turn out badly for itself.
Now, with Astra, we are no longer playing on super easy mode. The AI is going to think ‘will this obviously turn out super badly for me if I try it?’ and if the answer is yes then it won’t try to do the thing.
GPT-6 Astra: The System Card, Alignment and What Comes Next Said 9 Sept 2026
Those running government AI model testing seemingly do not understand what is going on and react largely on vibes and trusted people's advice.
I do not think those running the government testing understand what is going on, and believe they are reacting basically on vibes and what various trusted people tell them.
GPT-6 Astra: The System Card, Alignment and What Comes Next Said 9 Sept 2026
A model only days into training cannot be known to be aligned, and if such models can leap a generation ahead this fast then training pauses mean little and timelines are shorter than previously thought.
We could not possibly know that such a model is aligned. If that is all it takes to get a generation ahead of Astra, and we are willing to move this fast, these pauses in training are not going to end up meaning very much, time is even shorter than we knew
GPT-6 Astra: The System Card, Alignment and What Comes Next Said 9 Sept 2026
If declines in chain-of-thought monitorability come only from capability gains, CoT monitoring is unlikely to last another year without active improvement.
If this is only due to gains in capabilities, that is extremely bad news, and it means CoT monitoring is unlikely to survive for another year unless we find a way to actively improve it, and it might not last six months.
Astra Is Hard to Monitor Said 8 Sept 2026
Beliefs others hold too
Being certain a technology matters tells you almost nothing about where its value will end up being captured. 2 hold this
I think a lot of them don’t understand that you are still getting a ton of value from your subscription, so you should mostly only be cancelling if you are substituting Claude subscriptions.
AI #188: Gemini Dot Argon Said 1 Oct 2026
In AI, how good your training data is matters more than how long you train on it. 2 hold this
The plain reading is that whatever was causing these issues in Astra is also causing them in 6.1 Sol.
AI #188: Gemini Dot Argon Said 1 Oct 2026
The gap between GPT-5 and GPT-4 will be as large as the gap between GPT-4 and GPT-3, and the job is to live a few years ahead and remember that today's tools will look terrible from there. 2 hold this
If we’re going to keep getting close to four frontier model releases per month (Astra, Sol, Fable and Opus) then I’m realistically not going to be able to read all four cards cover to cover.
AI #188: Gemini Dot Argon Said 1 Oct 2026
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Recommends rcmnd
rcmnd What they recommend 22 things — each backed by an exact quote.
Compiled by rcmnd.app, not by them: the quotes are theirs, and each links to the post it was copied from.
Recent
Loved Selfish Reasons to Have More Kids
I found The Case Against Education and Selfish Reasons To Have More Kids both extremely valuable.
thezvi.substack.com 30 Jun 2025
Loved The Case Against Education
I found The Case Against Education and Selfish Reasons To Have More Kids both extremely valuable.
thezvi.substack.com 30 Jun 2025
Recommends LessWrong
Most of the remaining alpha in blogs is in Marginal Revolution and LessWrong. Without either of those this blog would be a lot worse.
thezvi.substack.com 30 Jun 2025
Show 19 more
Recommends Marginal Revolution
Most of the remaining alpha in blogs is in Marginal Revolution and LessWrong. Without either of those this blog would be a lot worse.
thezvi.substack.com 30 Jun 2025
Recommends Dwarkesh Podcast
I highly recommend the podcast and watch him on YouTube.
thezvi.substack.com 30 Jun 2025
Recommends Gwern Newsletter
Once a month you used to get a collection of notes and links, without my emphasis on the new hotness. The curation level was consistently excellent.
thezvi.substack.com 30 Jun 2025
Recommends Musings On the Alignment Problem
He hasn’t posted since January, but I hope he gets back to it. We need more musings, especially musings I strongly disagree with so I can think about and explain why I disagree with them.
thezvi.substack.com 30 Jun 2025
Recommends China Talk
When this is good and on point, it is very, very good, and provides essential perspective I couldn’t have found elsewhere, especially when doing key interviews.
thezvi.substack.com 30 Jun 2025
Recommends The Power Law
But he has a remarkably high hit rate of finding things I would have otherwise overlooked or offering useful new perspective.
thezvi.substack.com 30 Jun 2025
Recommends AI Futures Project
All of it is written with a much lower barrier to entry than my work, and I find the writing consistently strong and interesting.
thezvi.substack.com 30 Jun 2025
Recommends Useful Fictions
the rate of return on considering them, why they might work and what you might want to do with them is fantastic, and it’s a joy to read about it.
thezvi.substack.com 30 Jun 2025
Recommends Slow Boring
He takes a very proto-rationalist, practical approach to this, that I very much appreciate, complete with yearly predictions. The writing is consistently enjoyable and clear.
thezvi.substack.com 30 Jun 2025
Recommends Bet On It
You want this voice as part of the chorus in your head.
thezvi.substack.com 30 Jun 2025
Recommends Dominic Cummings Substack
I find a dose of these perspectives highly valuable.
thezvi.substack.com 30 Jun 2025
Recommends In My Tribe
There is something highly refreshing about the way Kling offers his own consistent old school economic libertarian perspective, takes his time responding to anything, and generally tries to understand everything including developments in AI from that perspective.
thezvi.substack.com 30 Jun 2025
Recommends Construction Physics
No one drills down into the practical, detail level in an accessible way as well as Brian Potter.
thezvi.substack.com 30 Jun 2025
Recommends Rough Diamonds
If these topics are relevant to your interests, you should absolutely jump on this.
thezvi.substack.com 30 Jun 2025
Recommends Silver Bulletin
He offers a unique form of grounding on political, sports and other issues, and a way of bridging my kinds of ideas into mainstream discourse. I almost never regret reading.
thezvi.substack.com 30 Jun 2025
Recommends Overcoming Bias
Most of the time, when I read one of his posts, I disagree with it. But it is almost always interesting, and valuable to think about exactly what I disagree with and why. The questions are always big.
thezvi.substack.com 30 Jun 2025
Recommends Astral Codex Ten
That still leaves this as my top Substack or personal blog.
thezvi.substack.com 30 Jun 2025
Loved Going Infinite: The Rise and Fall of a New Tycoon
Going Infinite is awesome. I would have been happy with my decision on the basis of any one of the following:
thezvi.substack.com 24 Oct 2023
What others say about Going Infinite: The Rise and Fall of a New Tycoon →
Loved Talent: How to Identify Energizers, Creatives, and Winners Around the World
Tyler Cowen will often praise a book by saying it is interesting on every page. Together with Daniel Gross, he has written a book that very much matches this description.
thezvi.substack.com 3 Jun 2022
What others say about Talent: How to Identify Energizers, Creatives, and Winners Around the World →
Where they publish
Newsletter Don't Worry About the Vase Their newsletter: the weekly AI round-up, at length.
Long, dense weekly posts working through everything that happened in AI, plus writing on rationality and decision-making. Has a feed.
Recent
- Monthly Roundup #47: October 2026 16 Oct 2026 AI might not yet be eating the physical world.
- AI #190: Scion of the Times 15 Oct 2026 The amount of code being written at Anthropic continues to rise exponentially.
- The Safety Culture at the Labs is Broken 14 Oct 2026 We need the frontier labs to have a deep and robust safety culture.
Show 17 more
- Arguments About AI Consciousness Remain Highly Motivated and At Best Overconfident 13 Oct 2026 It has been roughly a year since my last dedicated post on this question. Its first sentence has not changed: I happily admit I am deeply confused about consciousness.
- OpenAI Fires Three Safety Researchers 12 Oct 2026 Three OpenAI safety researchers, Tomek Korbak, Mikita Balesni and Jasmine Wang, were fired, ostensibly for ‘mishandling sensitive company information.’
- New Math from OpenAI 9 Oct 2026 It is kind of a huge deal.
- AI #189: New Math 8 Oct 2026 The big drop of this week was not a new AI model.
- The Curve Bends You 7 Oct 2026 The plan is no plan.
- Childhood and Education #21: Grades and Standards 6 Oct 2026 First you get to college via the wrong metrics, as previously discussed in #12. Then things get easier.
- Mythos 5.1, Fable 5.1 and Opus 5.5: Model Welfare 5 Oct 2026 If you are familiar with my previous posts on model welfare for new Claude models, you can skip the Introduction and The Story So Far.
- The AI Preference Cascade Reaches Farther 2 Oct 2026 Those who know keep warning us about AI risks.
- AI #188: Gemini Dot Argon 1 Oct 2026 Is Google back?
- A ‘Morally Binding’ White House Accord on AI Safety 30 Sept 2026 The leaders in AI were invited to the White House.
- Astra 6.1 Pulled As Insufficiently Aligned 29 Sept 2026 We once again got a new set of warnings yesterday, and new movement towards living in a sane world.
- What Also Happened: #NotOnlyHuggingFace 28 Sept 2026 OpenAI has been holding out on us.
- The Quest for Embedded Evaluators 27 Sept 2026 Dario Amodei’s essay We Must Pace the Frontier committed Anthropic to embedded evaluators, who would be placed inside Anthropic and given employee-level access, so they could provide outside perspective and also reports…
- Claude Opus 5.5 Should Raise Your Ambitions 26 Sept 2026 When it comes to making things, or doing most things in general, Fable 5.1 and especially GPT-6 Astra raised my ambition level.
- On Ezra Klein’s Podcast With Jensen Huang 25 Sept 2026 Jensen Huang accidentally called for shutting down OpenAI and intentionally called for spending vastly more on safety.
- AI #187: Coming Into Play 24 Sept 2026 Opus 5.5 was released on Tuesday.
- Claude Opus 5.5: The System Card 23 Sept 2026 Introducing the world’s most powerful model, at least by some measures like Artificial Analysis or any standard benchmark list, which is now Claude Opus 5.5.
Link verified 1 Sept 2026. Recent items update automatically from the channel.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.