Here are some links I consumed this month.

Culture

1. 9/11 in Berkeley

   This piece made me regret not being able to talk to Scott when I was at Inkhaven; I am even more ashamed that I have actually not read his blog before. My Inkhaven trip was poorly timed as I left a few hours after Scott arrived. I think no one should do Inkhaven like I did — remotely for almost half of it — because all the best mentors and guest talks happened in the last couple weeks.

2. How Podcasts Are Changing Comedy

   During COVID, I was taking lessons from Francis Foster, one half of Triggernometry and this was when I first learned about comedians hosting their own podcasts that weren’t genuinely big names (Marc Maron and Joe Rogan were already doing this). My brief time trying to standup ended with the realisation that most time is spent on the admin of scheduling newbie gigs, which leaves little time for actually writing material and practicing delivery (alongside having a real job).

   Now, I suspect a lot of early-career comedians have gone the route of podcasting/TikToks because they realised they enjoy having easier access to audience numbers that open mic gigs don’t allow for the one-time effort of setting up a “studio” (two devices for a zoom call are sufficient to start off). As one accrues an audience, they can scale up to something more professional. This can give you a vaster body of work than practicing standup does to create a lot more content in the environment you will perform in — much like blogging — which eventually helps you hone your craft. I think this explains the rise of podcasting more than the five ingredients the writer describes here.

AI

1. How employment is changing at Anthropic

   Comments here.

2. Post-AGI, we are all jobless aristocrats

Frontier Labs

1. Inside the Biggest Feud in Artificial Intelligence

   This was… not that great in terms of information density? It doesn’t paint anyone in particularly good light and points to a broader trend I’m noticing: we have entered the Hollywood-paparazzi-esque phase in the trajectory of founders in Silicon Valley, the makers of cultural tools; I suppose it is quite fitting that this is also playing out in California.

Science Impact

1. 2026 Conference on Physics and AI: Jascha Sohl-Dickstein

   Felt like such an important talk that I watched it twice (primarily because I didn’t know it existed in thread form): the first was at 2x at night then again, at variable speeds, the next morning right after I woke up. It presents many charts capturing the steep ascent of AI research progress, but does so in a far more optimistic manner — might be the speaker’s vibe — despite talking about what happens if people miss the moment we are in where humans still matter. One of the charts I am sitting with from the talk is from a two-year old Epoch AI report that tells us that we are still early; the talk is only 3-ish months old so I don’t know how to feel if one can still be early or is too late. But I’d err on the side of believing that it’s never too late to take on truly ambitious projects that are not yet on the exponential. More from me here.

2. The Age of Wonders and Terrors

   Simply excellent writing on what the Navier-Stokes work at OpenAI means for the maths community. This piece helped me see why Ilya Sutskever and Jan Leike wanted Scott at OpenAI because as he said: “they were fans of my blog”.

   Of course, I read this on the plane to SF and, upon landing, saw OpenAI’s new announcement of having solved even more major problems, one of which Alex Kontorovich has hailed as an instant Fields Medal.

3. Theory Beyond Theorems and Proofs: A Guest Post

   The last of the pieces I read on Scott Aaronson’s blog was not written by him, but by three other academics calling for the creation of a separate track for conceptual papers at leading conferences in their field of Theoretical Computer Science. This, they say, would deprioritize proofs, which is now the territory of AI. I am not sure if they are actually related but the ideas here reminded me of Michael Nielsen’s essay on vision papers.

4. Scott Aaronson on the Impact of AI on Science and Mathematics

   As you will notice from the links that follow, I went down a bit of a Scott Aaronson binge recently. I wanted to read his take (link below) on the Navier-Stokes/Millennium Prize work by OpenAI, but didn’t get around to it until I got on the flight to the Progress Conference. So, I listened to this as I was packing for SF. It was a bit of a whirlwind tour of his career and writing. So I ended up reading a bunch of other stuff on the plane as well that you’ll see in the links below.

5. My new course at UT Austin: AI Alignment Theory

   I’m now, in 2026, effectively teaching a course on Yudkowsky Studies. And it’s the most important course I can teach.

   Hilarious, but cool. I think it’s also cool that the academic ecosystem at UT Austin is flexible enough to permit running courses as discussion seminars. Gonna have to dig around to see if QMUL allows that.

Education Impact

1. What Shall I Teach My Son?

   Could’ve been under education too.

Risk

1. What I saw inside Anthropic

   At this point, I wonder where the newish issue on AI scares lies. This is not a dismissal of the piece — I like Kelsey’s flippant written voice — but Anthropic’s affable pursuit of free advertising by relieving NDAs worries me. It feels like the territorial advantage they’re trying to gain enables bio-risk, whereas OpenAI’s approach will enable intellectual disempowerment. Both seem like terrible niches to pursue.

AI Written Blogpost

1. Cognitive Mercantilism

   It took me a little while to read this essay, mostly because I liked it and began digging around for VGR’s prompt as I recall his announcement of using AI to write his post from hereon. You can see his prompt in the next item. This essay is, I think, a reminder that AI writing in the right hands and harnesses could become more than readable in collaboration with a human for nonfiction (but maybe more?) that it feels like a compelling read. But it also might make you feel like my friend, who said the following:

   Is he saying that the prompt that he put on there is all he wrote and the rest was chatGPT?! In that case, the blogging game is gone, maybe I shouldn’t consider blogging anymore!

2. VGR’s prompt

   Here we go with VGR’s prompt. My suspicion is there is a custom harness though it might also be a fine-tuned model? If it’s just a back and forth with GPT, then it is wild because for a post that Pangram says is ~70% AI — not AI-assisted, but AI! — this is really quite good.

Jobs

1. Hunger beats tenure: Advice for junior engineers

   Great news! Maybe I’ll finally be able to get a job at a frontier lab.

Governance

1. What Should the Rules for A.I. Companies Actually Be? We Asked the Experts.

   A very powerful star cast of thinkers on AI, mostly on the non-technical side, which is probably what it should be when thinking about the societal impacts of this technology. Much of the conversation centred around regulating the AI companies and the role of America in doing that while also negotiating with China. But the other way to think about this conversation that it covered three types of risks: catastrophic, existential, and imminent.

   If one were to think of the conversation as between two parties, one for the progress of AI and one to pause or even stop development altogether, then there are good arguments on both sides, and I think grand assumptions in either case. It’s just as difficult to predict whether it’s going to go well as it is to predict that it will go badly, or at least, to predict that it will go badly, I mean.

   So I think we should be trying to do the thing that allows us to inch forward in a way that makes sense. But ask me on another day and I’m pretty sure I’ll change my mind. It really depends on the Kool-Aid I’ve been drinking that day, and it tells you how difficult it is to convince someone like me that one side is more right than the other.

Philosophy

1. Do not conquer what you cannot defend

   This LW post is hard to categorise as it probably fits more under corporate ideology more than philosophy, but it’s early days for my linkthreads so forgive me. It talks about organising around rationalism and EA and jumps between presenting them in ways that feel like shades of a religion/cult and a startup. I like to joke that EA feels like Buddhism with numbers to me. They seem to obsess over performance measures and deploying capital in specific areas that make it seem corporate-like (this is not necessarily a knock on them, but it is a limited worldview) while having the Buddhist objective of reducing suffering/doing good.

Inkhaven

Inkhaven Archives

1. Omniscient Entities

   Linked to the post right above this one; not sure the concept of “omniscience” fits here and neither do the examples. Also, I think the chart is woefully under-explained because a lot of what is being described relies on models of something being either decentralised or centralised.

2. The Anti-Straussian Ceiling

   If I understood this correctly, the author is saying that some wordcels (which spans written and spoken) begin as edgelords. This helps them grow their social clout online, but eventually become too successful to sustain this edginess. In some cases, it might be that they wish to remain popular over being provocative. Or, in some cases, they start institutions which are the meta-way of betting on their edgy view without saying it. Over time, they speak in a more meta and non-committal manner as it allows them to adapt to whatever is palatable in the moment’s political climate or aligned with their organisation’s objective (even if it is at odds with their personal opinion). Some of this is true, but I think it misses that people’s world models get richer over time through more information and, more importantly, experience; this should make people hedge a little more or just say they are unsure. I don’t see a problem with this as updating based on information—being Bayesian—seems reasonable.

   What this doesn’t fully explain to me is where Tucker Carlson sits. Like why did he say any of the totally incorrect things he usually does (the latest being where he gets schooled by Jacob Rees-Mogg)? This trend is not limited to just him, but other popular individuals who operate as media forces like Andrew Tate, Joe Rogan, etc.

Application Tips

1. Your First Writing Sample Is The Worst One

   Not so sure about this claim, but since I don’t make the rules, I’ll just leave this here as a potentially useful rule-of-thumb for anyone applying to Inkhaven 4.

2. Gatekeep, Gaslight, Boyboss

   I just enjoyed reading this from the head o’ the show this time around at Inkhaven 4.

Space

1. A Call to Action

Technology

1. Foster + Partners takes design to the moon

   These architects have been thinking about lunar bases from the UK for some time. Good to know.

2. How SpaceX Streamlined the Raptor Engine

   This is such a good practical article on rockets! I feel like I need to revisit my own post on rockets; such a beautiful line:

   Now instead of just using the energy stored in the pressurized propellants, we’re also using the energy stored in their chemical bonds, and converting that energy into thrust.

   We shouldn’t be surprised that SpaceX doesn’t explain how their engine designs were optimised. But this is wild:

   This streamlining has come alongside meaningful gains in performance, with the Raptor 3 providing about 35% more thrust than the Raptor 1.

   How does one achieve these results by elegantly simplifying a design? On the surface, at least, but the subsurface engineering would be even crazier which Brian confirms:

   Another substantial change on Raptor 3 is that much of the piping hasn’t been eliminated, but moved internally, by way of 3D printing many of the components.

   In real physics, you can more often hide a part than eliminate it, but 3D printing does both: a pipe and its joints become a channel inside one printed piece. Delete the part, as Elon says, and keep its job.

3. Superpowers Race to Put Nuclear Reactors on the Moon

   NASA wants contractors to build and launch a lunar nuclear reactor by December 2030 because Russia and China are aiming to do so by 2036. American companies making expensive small microreactors (i.e., generating up to 10 megawatts of electricity) could benefit here from NASA technology development awards. NASA plans to use nuclear power to propel a spacecraft to Mars in December 2028.

   However, a snag is that Russia controls the largest supply of high-assay low-enriched uranium (HALEU), considered the safest nuclear fuel for space missions and in short supply in the United States.

   Technical challenges comprise: landing heavy, potentially radioactive material onto bumpy lunar terrain remains unsolved; operating power systems near the moon’s south pole that must survive temperature swings between 130 and -334 degrees Fahrenheit; and lunar nights stretching over two weeks of Earth time limiting reliance on solar.

4. Sunlight at night: US startup’s demonstration of lighting Earth from space draws criticism

   A piece on Reflect Orbital’s demonstration of their tech using floodlights near Mount Whitney. Locals were unhappy with the tests: some described the lights as “blinding.”

   Tinkering with the natural world always upsets some people, while others play up the benefits. Astronomers will object to anything that obscures their view, and other longer-term concerns include disruption to human circadian rhythms, possible eye damage akin to looking directly at a solar eclipse, and effects on wildlife. Reflect Orbital presents its technology as an environmental benefit that reduces reliance on fossil fuels: beaming sunlight onto solar farms around dusk and dawn would extend the hours they generate power. Other proposed uses include bathing crops in sunlight, replacing broken streetlights, powering military operations, and helping search-and-rescue teams and other emergency responders at night.

Science

1. Who Owns the Moon?

   Linked to the above piece, here are some interesting bits on the treaties:

   The Outer Space Treaty declared that countries could not claim outer space “by means of use or occupation.” Space exploration, the treaty said, should benefit all countries. (The treaty was focused on arms control, and it barred nuclear weapons in space.)

   In the late 1970s, the Moon Treaty declared lunar resources a common heritage of humankind and insisted upon international governance of their exploitation.

   Only 17 countries agreed to the treaty but the United States, China, and the Soviet Union did not back then; I suspect this still applies to the split up USSR states (including Russia). A 2015 American law allows companies to mine in space and the non-binding principles of the Artemis Accords declare that “extracting resources from the moon, Mars, comets or asteroids does not violate the Outer Space Treaty”.

   This is not new for much of the space community, but it is possible that the AI race dynamics could ratchet things up to make these things pertinent quite soon.

2. Worlds More Habitable Than Earth? They’re Out There.

   I became less supportive of space astronomy missions ever since I learned about JWST’s delays and budget exploding by 10x. Space telescopes are luxury Instagram projects needing a more cost-effective and time-optimal solution; Cosmic Frontier seems to be working on this. I’m unsure if these newer, cheaper satellites can support the identification of superhabitable worlds. Excerpts from the article:

   photosynthesis, the process that supplies the energy for Earth’s entire food web, can happen on planets orbiting these orange stars.

   Scientists call these “superhabitable worlds,” planets that would allow carbon-based life to persist longer and become even more complex and diverse than on Earth.

   For Dr. Heller, the hunt for more habitable worlds is about “finding out if we are alone, if we could have been found, if there’s someone to talk to,” he said.

   “Although I am not, at least in my everyday life, a very spiritual person,” he added, “it’s nevertheless something that at least drives me — this question, Could there be someone?”

Education

1. UK universities cut jobs, research and teaching amid squeeze in overseas students

   Lucrative overseas students, who have helped universities subsidise UK students, fell by 10 per cent in 2025-26.

   On Monday an annual survey by Universities UK, the umbrella body for higher education, found 38 per cent of the third of members that replied were carrying out compulsory redundancies, up from 11 per cent in 2024.

   Some 79 per cent said they were making voluntary redundancies, the first time the question has been asked.

Technology

1. How Might We Learn?

   Unwittingly, I’ve been incorporating elements of Andy’s thoughts on learning into my own approaches for self-learning, but have now begun integrating aspects of this into how I teach with LLMs in the classroom. It is an ongoing experiment that I hope I will write about soon-ish.

Cinema

Digger

1. ‘Digger,’ Tom Cruise’s Big Flop, Is a Secret Masterpiece

   This piece in defence of Digger — a movie I found abhorrent — tries to explain its genius. My problem was the movie’s ineffective use of tropes for the woke and powerful in explaining our collective inaction on the climate crisis. My Digger-disdain might be explained in my watching the excellent Nathan Fielder over the last couple weeks, who explores similar themes in The Curse. It offers a more interesting examination of themes and people without trying too hard to be as meta about it, while using similar mechanics of using narrative within a narrative.

2. The Digger disaster: what an epic movie flop means for Tom Cruise and Hollywood

   I think this is fair to say:

   But Digger is unique: an unprecedented folly of misjudgment, bad timing and spectacular hubris.

London Film Festival

1. Possible Love review: Lee Chang-dong returns with haunting masterpiece

   “Possible Love” is as dense as our best literature, filled with rich character beats that feel of a piece with what came before and where these characters are going next.

Cinema Reviews

Digger

1. Tom Cruise’s Digger has flopped spectacularly – but I think it is a misunderstood masterpiece

   I’ve been reading defences of Digger to try to see if I missed something deeper, but I am yet to come across anything satisfying. That so many, including GDT, have Digger psychosis is a real phenomenon. Take for example what this Guardian piece says:

   And, if we are honest with ourselves, do we just want art that tells us the truth, or reinforces what we already think? Iñárritu’s answer is the latter; we’d probably settle for a middling play with a bad title if it gave us two hours of distraction and someone to blame.

   Well, I hate to say it, but if Iñárritu was right, then the majority would have enjoyed the movie, but clearly they found it to be a bloviating hunk o’…

London Film Festival

1. Fatherland review – Sandra Hüller brings a bayonet of intelligence to Paweł Pawlikowski’s taut return

   Though it is clearly a very good movie, I disagree with Peter Bradshaw’s assessment of Fatherland being a perfect movie.

2. ‘All of a Sudden’ Review: Virginie Efira and Tao Okamoto Forge a Life-Changing Friendship in Ryusuke Hamaguchi’s Life-Giving Drama

   A runtime of three hours is quite a bit for any movie, but I genuinely didn’t feel duration in this case. This is quite incredible as I came to this screen from having finished the newly restored Pakeezah only 15 minutes earlier; including its prescreening chat, that was another three hours.

   Much of the way the movie is shot and its melancholic beat reminded me of Perfect Days. Of course, the Japanese also helps. The back-and-forth switching between French and Japanese may seem pretentious to some, but I felt it works well here.

   All of a Sudden’s weakest moment is where the leads engage on the ideals of capitalism, which provides the pivot for the movie. I’m glad that this didn’t last long and it returns to a tender portrayal of loss (of memory and life). There were many moments that left me with moist eyes, but I do get emotional quite easily with such content.

3. Minotaur: an elegant and sinister Chabrol adaptation

   Spectacular! One of the best movies I have seen this year and a well-earned 98% fresh on Rotten Tomatoes. I will have to watch La Femme infidèle that it’s an adaptation of.

4. Film Heritage Foundation’s 4K Restoration of Kamal Amrohi’s Masterpiece “Pakeezah” (1972) to have its world premiere at the prestigious Festival Lumière in Lyon, France followed by the UK premiere at the London Film Festival

   Bollywood movies definitely had something about them that has been lost in the last couple of decades. Sholay and Lagaan are some of the best examples of Bollywood at its best; I’ve now added Pakeezah to the list.

   The making of the film has all the romance, drama and heartbreak of the film itself. “Pakeezah” released on February 4, 1972 but it’s journey to the big screen was a painstaking journey of fifteen years from the time filming began in the mid – 1950s.

   The challenge of restoring it in 4k must have been one heckuva task:

   With no original camera negative and no surviving cast and crew of the film, ensuring that we matched the original colours of the film that had been so integral to Kamal Amrohi’s vision was a challenge. Through our research we were able to find original shooting stills of the film. We also regularly consulted with Tajdar and Rukhsaar Amrohi, the son and daughter of Kamal Amrohi, who had vivid memories of the film and had also been present during the shooting of the film.

5. Possible Love review – a tale of two couples is finely judged study of exploitation and friendship

   Watched this very good movie at the London Film Festival, in the haze of a post-Progress Conference jet lag.

6. Fatherland review: A sparse but powerful piece of heartbreaking historical fiction

   Fatherland was good in ways most good movies on the Second World War are good: great performances in black-and-white presentation that treat the subject matter seriously of what was an objectively horrible moment in history. Outstanding camerawork throughout this movie helps frame a tale of a somewhat self-serving father caught in no man’s land — in more ways than one. Of course, the addition of Sandra Hüller is becoming another marker of high-quality Holocaust-era movies.

   Writing this comment, also made me realise that Quentin Tarantino remains the only filmmaker (that I can recall) who transformed this sensitive subject matter into a dark comedy.

7. Klara and the Sun review – Taika Waititi’s manic pixie dream robot sci-fi is a sentimental mess

   Peter Bradshaw very generously awarded it a solitary star. I regret sitting through this movie even more than Digger. I was totally dumbstruck by the applause at the end of the movie that has made me lose respect for the crowd that attends the London Film Festival.

   The novel was good, not great, but I see no reason for why the movie deviates from its source as much as it does, in its story and general tone, in the worst ways possible. The result is a poorly executed Tim Burton-like movie that simply never takes off at any point.

   Bradshaw’s suggestion to watch After Yang is excellent; my time would have been better spent rewatching that.

8. ‘I’d consider replacing one of my children’: Taika Waititi on robot kids, his cheese-o-meter and Klara and the Sun

   Let it be known that “young influencers” — whatever that means — are tasteless.

   Bradshaw and others are right and other views are very, very wrong.

The Devils

1. The Devils movie review & film summary

   A friend spotted an unread copy of Huxley’s Devils of Loudon on my bookshelf and gifted me a ticket. I loved it and I don’t get Roger Ebert panning the movie!

   Gruesome and odd, but with grand sets. And, man, I loved Oliver Reed in Gladiator, but his embodiment of Grandier is mesmeric! The man has such presence! His prowess alongside the weird and weak performances of most others in it makes the movie feel like satire even in spots it might not wish to be.

2. The Devils review – Ken Russell’s 1971 possessed-nuns tale is a frenzied aria of confrontation and hysteria

   This is, of course, the correct take by Peter Bradshaw.

Sport

Cricket

1. Bavuma’s nine-hour masterpiece was art we risk losing in Test cricket’s great purge

   I think this kinda writing is annoying and does more to make people forget about Bavuma’s innings than anything else.

Romance

1. living to know

   Probably the only thing I’m gonna read on the topic of romance this month (or at least share). Romance deserves its own section just to remind me that there’s more going on outside AI (sounds odd to read just as much as it is for me to write it). Someday, I will lean into “not being a typical man” and write more commentary on the matter, but that is hard to do. One cannot write such things without revealing private matters and my blog is read by students and colleagues. So, there are matters that must be left private.

   But, until the day I reveal more, I will say I don’t buy into this idea of “emotional extremism”, which is Ava’s phrase for being direct about your expectations, as being beneficial to everyone. The risk of being too available to others could turn into being too demanding of others, but I have seen that people do more of one than the other.

Podcasts

1. That Daily Brain Rot Break Is Taking Us Back to the Dark Ages

   Didn’t know someone had said that we will get the Homer of Tik-Tok — a hell of a thing to say, but we do have the Homer of Cinema in Nolan, right? Someone needs to write an anti-Fahrenheit 451 that imagines a dystopian future where we have hyperliteracy, a world where people think we are reading too much.

2. Neal Stephenson On AI And What Comes Next

   Fun listen; hadn’t heard about Whenere, that is some kind of movie script editor where users write stories with AI in some manner. This can then, I think, either be watched (or even played?) inside Fortnite; this bit is enabled by Epic’s backing. It sounds like they are starting with bringing Pride and Prejudice in as a first project; I suspect P&P is in the public commons now, but Neal’s books are tied up in Hollywood contracts that he can’t get out of so they can’t begin with his works. While he thinks of Whenere as a better, more humane tech, I am less confident. It can, like most things, tickle the human senses in a way social media does where there are some creators and more viewers. Nonetheless, it sounds like a novel kind of content creation process that allows more participation and should help more people feel creative via AI nudges. I’ll probably read the linked Substack post in detail sometime in the future.

3. Robert Caro – The master biographer reflects at 91

   It’s just heartwarming to repeatedly hear Caro tell Dwarkesh how good his questions are. At the same time, while the questions were unique, I noticed that it is hard for Caro to add new details. But I think the best part about the interview was the way Dwarkesh found connections and phrased them very gently: the similarities between the personalities he wrote about and himself, whether it was his dogged persistence to get to the very bottom of everything he wanted to get to, or the loneliness of pursuing his craft.

Miscellaneous

1. Every Bay Area House Party

   Read this in preparation of Mikhail’s London Bay Area House Party house party. Fun stuff. Having watched Ken Russell’s The Devils right before the party, I pretended to be the new founder of an Antichrist cult leader as I was quite sure that Peter Thiel would want to fund this (read Scott’s essay — it’s hilarious!).

2. Minotaur review: French thriller becomes searing Russian commentary

   Gleb, as the CEO of a logistics company, is an important man in this small Russian city. But his firm is haemorrhaging staff even before he is ordered to draw up a list of 14 more employees to be forcibly drafted and sent to fight, perhaps to die, in Ukraine. The Minotaur of ancient myth required the periodic sacrifice of 14 innocents.

3. Why Washing Machines Took Decades to Catch On

   People who could afford laundry services could avoid the time and effort home washers still demanded. In the South, racism kept Black laundresses’ labour cheap and made washing seem beneath white women’s status; that establishes neither that outsourcing cost less than owning a machine nor that families rejected machines at any price.

4. Against supplements - Works in Progress Magazine

   I’m not a supplements guy myself so I approve.

Here are some other things I skimmed:

1. That Alien Message

In all, I consumed at least 54 articles and at least 148,025 words this month.