Greg Egan

I am a science fiction writer and computer programmer.
Latest novel: MORPHOTROPHIC.
Latest collection: SLEEP AND THE SOUL.
Web site: gregegan.net

Web site: https://www.gregegan.net
Location: Perth, Western Australia

Public Key

npub12cuzzqzv8e8y7na77a9zv47mx2v6pec60qs5h7takzmv5p0mw0fsmgd38p

NIP-05 Address

[email protected]

Profile Code

nprofile1qqs9vwppqpxrunj0f7l0wj3x2ldn9xdquud8sg2tl97mpdk2qhah85cpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dspthf7h

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub12cuzzqzv8e8y7na77a9zv47mx2v6pec60qs5h7takzmv5p0mw0fsmgd38p

Show more details

Last Notes

2026-07-16 02:56:02 UTC

Anything meant to stick to human skin should be rated by its ability to cope with the existence of body hair, on a scale from:
0 for Band-Aid Aquablock: What! You put me on your *calf*? All I can do is fall to the ground.
to:
8 for Leukosilk: Ah, mildly hirsute here, no problemo!

2026-07-14 08:10:56 UTC

Lymphorrhoea is extremely disgusting — and comes with the added frisson that it makes me feel like I’ve been transformed into a character from my novel MORPHOTROPHIC, set in an alternative version of Earth where people’s cells sometimes say “Fuck this body, I’m going to run off and make my own life as a free agent!”

2026-06-27 13:42:07 UTC

My phone acquired a handful of odd bugs yesterday (unable to use mobile data, 3-button navigation not working, quick settings panel limited to one row) after being present [switched off] in the room where I had a CAT scan.
A factory reset got everything back to normal. I suppose it could be a coincidence, but if it wasn’t, I hope all the computing hardware crucial to the scanner itself was better shielded against whatever spillage corrupted the phone's flash storage.

2026-06-26 03:38:06 UTC

Lots of regrettable silliness here: two papers by Max Planck (!) have been retracted, probably by some automated process that based the decision on the same material appearing in another journal (which was far more common at the time, and even Einstein did it) and a rebuttal to a work being published under the same title (but not author) as the thing being rebutted.
But the funniest line in the article is:
“Springer Nature is nevertheless still selling the empty PDF for $39.95.”
https://www.science.org/content/article/why-have-papers-one-history-s-most-famous-physicists-been-retracted

2026-06-19 07:54:41 UTC

All the eBooks of my work that I self-publish are DRM-free, including the ones from Amazon, but Amazon have only recently added an option to allow their customers to download these books as ePubs or PDFs, as well as the special format used by Klndle readers.
It’s up to the individual publisher of any eBook on Amazon to decide:
(a) if they want the book sold with or without DRM
(b) if DRM-free, whether or not to allow downloads as ePubs / PDFs, which can be more easily read on non-Kindle devices.
I’ve opted for DRM-free, any format.
Amazon don’t make these choices easy for their customers to discover. DRM-free books for the Kindle used to have the description “Simultaneous Device Usage: Unlimited”, which was already rather cryptic, but now even that seems to have vanished.
Links to all my self-published eBooks. [If you don’t like Amazon, regardless of DRM status and file format, these are all available from many other vendors.]
https://www.gregegan.net/BIBLIOGRAPHY/Ebooks.html

2026-06-18 10:42:36 UTC

I think this is well worth reading. I've been a bit frustrated by some over-confident claims that data centres in space are literally impossible due to the cooling issues alone, and other claims that those issues are fairly minor. I am nowhere near close to being familiar with all the real-world engineering nitty gritty; I understand a fair bit of thermodynamics, but that is not enough to really settle things. FWIW, the arguments here seem well grounded in physics and the technological/economic/legal issues raised strike me as plausible.
https://robtow.substack.com/p/spacex-ipo-orbital-data-centers-and

2026-06-04 04:37:38 UTC

NO, ARTIFICIAL INTELLIGENCE IS NOT CONSCIOUS
Taken to its logical conclusion, this line of thinking is absurd—and damning.
By Ted Chiang
https://www.theatlantic.com/philosophy/2026/06/no-artificial-intelligence-is-not-conscious/687378/

2026-06-02 06:04:32 UTC

“… paint damage, floor damage, damage to a kitchen doorframe, bent poles in dishwasher racks, water damage and scratches on wooden credenzas … decorative items and books moved from shelves to drawers … ”
Airbnb hosts need to add a new clause to their agreements: no robots.
https://arstechnica.com/ai/2026/06/allegedly-trashing-airbnbs-to-test-robots-puts-startup-in-legal-trouble/

2026-06-02 05:23:46 UTC

https://www.quantamagazine.org/key-chemistry-question-answered-no-quantum-computer-required-20260529/
A fascinating article in Quanta on work to find the ground state of a key molecule in nitrogen fixation via an ingenious classical computation — despite this problem often being cited as a potential benchmark for quantum computers.
We’re in the middle of a fertiliser supply shortage that is likely to cause a crisis in food security for millions of people around the planet. I wish there was more work being done on incorporating biological nitrogen fixation into more crops, whether through genetic changes to the crops themselves, or through nitrogen-fixing bacteria that can work with more plants.
Coincidentally, I recently heard a Radio Lab program on the life of Fritz (industrial nitrogen fixation) Haber, and yeow, after the great famine-busting triumph, it was dark, dark, dark.
https://radiolab.org/podcast/180132-how-do-you-solve-problem-fritz-haber

2026-05-31 07:58:42 UTC

- reply

Here is the collection of 7 cones for the tetrahedra on its own (the cones themselves are infinite, but we draw them as finite square or triangular pyramids.)
Omar says Benoît Bertrand and Lucía López de Medrano don’t believe this is a new result, but they have not been able to find a reference to it.
https://mathstodon.xyz/@oantolin/116656345048312313
https://media.mathstodon.xyz/media_attachments/files/116/668/123/690/187/269/original/4c746db4004e787c.mp4

2026-05-31 07:56:58 UTC

How can we generalise the fact that the angles at the vertices of a triangle add up to 180°, to apply to tetrahedra and higher-dimensional n-simplexes?
Here’s a result discovered by Benoît Bertrand and Lucía López de Medrano that I learned from Omar Antolín.
Pick one vertex of the n-simplex, colour it blue, then consider all 2^n-1 ways to colour the other vertices red or blue, omitting the all-blue colouring.
For each such colouring, collect all the vectors that point from a blue vertex to a red vertex, and form their cone: the set of linear combinations of these vectors with positive coefficients.
Then these 2^n-1 cones will all fit together to form a half-space!
This is what we get for triangles.
https://media.mathstodon.xyz/media_attachments/files/116/668/117/946/686/162/original/b2f82894d9e08687.png

2026-05-30 22:48:42 UTC

- reply

What’s cool here is that you are correlating the position state of each photon with its polarisation state, rather than measuring the position in way that irreversibly couples it with the environment.
After doing that, if the polarisation is inaccessible to you, you need to trace over it to get the effective density matrix, which turns it from a pure state into a mixture of position states.
But the polarisation isn’t really inaccessible, so the diagonal polariser can interact with it and transform the state back into a pure state with no definite position state.

2026-05-28 08:09:12 UTC

- reply

Now I feel a strange compulsion to write a story in which there is a zombie outbreak in the UK along the lines of the “28 ... ” movies, but it is mostly covered up by the authorities, except that it degrades various services, e.g. trains are thrown into disarray by the need to scrape crushed zombies off the tracks.

2026-05-27 05:25:43 UTC

Most likely, no human typed the word “permission” here. Rather, they misspelled “persimmon” in such a way that autocorrect considered “permission” the closest valid word.
Pure spellchecking that *flagged* the typo would have alerted the user, who would have made the right choice.
Spellcheckers are great, but no serious writer/translator/copy-editor should enable autocorrect (let alone software that meddles in even more complex and unpredictable ways).
https://media.mathstodon.xyz/media_attachments/files/116/644/870/506/166/076/original/b59954bd9dad2326.jpg

2026-05-22 07:19:48 UTC

- reply

Just *projecting* from a high-dimensional lattice won’t give you a quasicrystal, there’s a certain kind of filtering of the points required.
https://www.gregegan.net/APPLETS/12/deBruijnNotes.html

2026-05-20 23:22:01 UTC

This is certainly an interesting and important result in geometry, in and of itself!
Several professional mathematicians quoted here comment on the significance for the practice of mathematics of the fact that this was found by an LLM.
FWIW, I’m not hugely surprised that there are *some* interesting and important results that can be found this way — and once we make more use of formalisation there will be a whole lot more, with a lot less errors and slop to wade through — but I think it’s still likely that plenty of scope will remain for humans settling similar conjectures that software is unable to resolve, just as certain smart, tenacious humans have always found proofs that other smart, tenacious humans worked on for decades and were unable to resolve.
https://openai.com/index/model-disproves-discrete-geometry-conjecture/

2026-05-12 23:55:02 UTC

- reply

Here’s a piece of a quasicrystal with icosahedral symmetry.
I describe how to get this with the 3D version of the cut-and-project method here:
https://www.gregegan.net/APPLETS/12/deBruijnNotes.html#HD
https://media.mathstodon.xyz/media_attachments/files/116/564/293/121/447/024/original/c82f0c9f8a2ac60a.mp4

2026-05-10 07:41:10 UTC

My local library shelves its fiction by genre (Crime, Historical, F&SF, ‘Relationships’), a practice I loathe, but I’m morbidly curious to learn if anyone else’s library (a) does this, but (b) has also coined some bizarre, completely nonstandard name for those few books they’ve decided can’t be squeezed into any genre at all.
My local library has chosen “Allsorts”. Gaaah.
[I suppose they decided that “Literary Fiction” sounded too snobbish, and on that I suppose I agree with them, but this weird, infantile neologism goes to the opposite extreme. Just shelve everything alphabetically by author, put colour-coded genre indicators on the spine if you must, but let people slip serendipitously out of their silos by browsing the whole damn gamut.]

2026-05-10 04:17:04 UTC

“A study that claimed OpenAI’s ChatGPT can positively impact student learning has been retracted nearly one year after publication. The journal publisher, Springer Nature, cited “discrepancies” in the analysis and a lack of confidence in the conclusions—but not before the paper racked up hundreds of citations and made the rounds on social media.”
https://arstechnica.com/ai/2026/05/influential-study-touting-chatgpt-in-education-retracted-over-red-flags/

2026-05-09 14:13:41 UTC

- reply

It’s pattern matching that would allow it to discern what proofs from the literature are “relevant,” and what recomposition of parts of them “solve the problem”.

2026-05-09 12:03:26 UTC

- reply

In previous cases of automating mathematical activities, we’ve usually known very explicitly exactly what we were doing: computing trig tables, enumerating and checking graphs for the 4-colour map theorem, etc. This makes it relatively easy to understand the time and memory resources required, and how far they can be scaled up before hitting computational complexity limits.
It’s much less clear what LLMs are doing when arriving at these kinds of proofs, but it would be helpful if the people building and studying LLMs could make it equally transparent.
My guess is that the “algorithm” being effectively performed here (though of course not explicitly programmed as such) is something roughly along the lines of:
• extract a set P of relevant / related proofs from the literature
• break P down into a set of suitably fine-grained steps, S
• iterate through subsets of S looking for any collection of steps that will solve the problem
It’s impressive that it’s possible to automate these three things at all, even as imperfectly as LLMs do them, but if this description is more or less correct then it will eventually hit the same kind of scaling issues as any other combinatorial approach.
It might be possible to make this kind of smart-brute-force search more smart and less brutish in the manner of Alpha Go etc., but strategically pruning the decision trees for assembling valid proofs is likely to be much tougher than it was for Go.

2026-05-05 21:56:39 UTC

Breaking: Rubio, Araghchi announce negotiations to divide up parallel worlds between those in which US, Iranian conditions are met to each party’s complete satisfaction.

2026-05-05 06:04:51 UTC

I rarely used to weigh myself, but I recently started doing so daily, and I was surprised by how much the results fluctuated.
In this study:
https://pmc.ncbi.nlm.nih.gov/articles/PMC7192384/
they found:
“Within-week fluctuations of 0.35% were observed, characterised by weekend weight gain and weekday reduction which differed between all groups.”
But I seem to have peak-to-peak fluctuations closer to 4%, or ten times as much as the 0.35% seen in that study!
My food intake and exercise patterns are pretty constant, apart from doing a longish run every Sunday.
I’m curious as to what kind of fluctuations other people have measured on this timescale — especially when they’re not knowingly changing any of the relevant factors.
https://media.mathstodon.xyz/media_attachments/files/116/520/448/461/021/767/original/fcc5c12d7c6ecb20.png

2026-04-27 00:02:56 UTC

“An AI Agent Just Destroyed Our Production Data. It Confessed in Writing.”
https://x.com/lifeof_jer/status/2048103471019434248

2026-04-25 14:58:12 UTC

- reply

Sorry, I don’t recall. I am more likely to have argued about evolving an intelligent entity rather than training one (as per https://www.gregegan.net/MISC/CRYSTAL/Crystal.html ) but I have no memory of a Usenet post on the subject.

2026-04-23 11:05:09 UTC

A great story on how the Finnish “low-temp mafia” and others are working to refine and replace the bulky, expensive He-3 dilution refrigerators needed for millikelvin quantum computing.
https://www.science.org/content/article/helium-3-runs-scarce-researchers-seek-new-ways-chill-quantum-computers

2026-04-21 08:09:03 UTC

- reply

FWIW, it seems possible that it was once spelled “cue”.
cue | kjuː |
noun
a long straight tapering wooden rod for striking the ball in snooker, billiards, etc..
origin
mid 18th century (denoting a long plait or pigtail): variant of queue.

2026-04-19 02:42:55 UTC

Collaboratively solving a word puzzle with a relative yesterday:
They: Ruth.
Me: That’s not a word!
They: You can be ruthless, so …
Me: That makes sense, but English is not that logical!
Concise dictionary: No “ruth”.
Complete dictionary: Archaic now, but yes, “ruth”: a feeling of pity, distress, or grief.

2026-04-09 03:18:36 UTC

Multiple lensed images of supernovae help pin down the Hubble constant:
“If SN Requiem reappears this year, it will lead to a higher value of the Hubble constant and validate the astronomers’ camp. If the supernova is not seen until next year, it will lead to a lower value…”
It’s astonishing that we can (sometimes) see the same supernova *billions* of light years away along multiple gravitationally-lensed paths that produce time differences on the human scale of 10 years or so.
https://www.science.org/content/article/how-fast-universe-expanding-cosmic-illusions-may-hold-answer

2026-04-08 21:39:55 UTC

- reply

Not sure if you’re joking, but the book’s title is a bit of a giveaway.
https://en.wikipedia.org/wiki/Parallel_transport#Approximation:_Schild's_ladder

2026-04-08 21:34:14 UTC

- reply

Parallel transport?

2026-04-08 09:37:16 UTC

- reply

It’s odd that neither that page, nor the linked page on spherical excess, spells out the connection between spherical excess and the *area* of the spherical triangle.

2026-04-02 11:56:34 UTC

I haven’t eaten liquorice for ages, but when I read “You should not eat large amounts of liquorice while taking this medication” …
(a) I had a sudden craving for liquorice
(b) it turns out probably *no one* should eat *large* amounts:
https://pmc.ncbi.nlm.nih.gov/articles/PMC8962506/

2026-04-01 14:03:30 UTC

- reply

No, that’s exactly wrong. What you’re saying here applies to spacelike geodesics. Every ordinary object’s world line is a timelike geodesic; no object’s world line is a spacelike geodesic, as that would mean travelling faster than light.

2026-04-01 13:19:17 UTC

- reply

“When you say 2+1-dimensional is that a 2D plane plus time? so the model of our universe would need to be (3+1)-dimensional? Is this flatland stuff?”
Yes, I’m starting with “flatland” (2D space plus time), and then carving out a hyperboloid within it that is even lower-dimensional, just 1D space plus time.
So this (1+1)-dimensional spacetime is the simplest spacetime with constant curvature, a bit like the 2D surface of a sphere is the simplest space with constant curvature.
But what’s cool is that, while the geodesics on a sphere behave the same way whichever direction they point — a bunch of parallel geodesics will always converge — the geodesics in (1+1)-dimensional de Sitter spacetime behave in three different ways! If they point in a “spacelike” direction [more spacey than timey] they converge. If they point in a “timelike” direction [more timey than spacey] they diverge.
And if they point in a “lightlike” direction [equal parts spacey and timey, at least in units where c=1] they neither grow closer nor further apart.

2026-04-01 13:11:51 UTC

- reply

A timelike geodesic is the world line for a free-falling object. In the absence of gravity, spacetime is flat (that is Minkowski spacetime, the spacetime equivalent of flat Euclidean space) and the world line of a free-falling object is just a straight line through spacetime, which we would see as the object moving along a straight line in space at a uniform velocity.
But where there is gravity, spacetime is curved, so just as the geodesics on a curved surface are not straight lines, the geodesics through spacetime are no longer straight lines.
For example, the timelike geodesic of the Earth as it orbits the Sun is a kind of elliptical helix through spacetime, as it follows an ellipse in space while its time coordinate increases.

2026-04-01 12:53:00 UTC

- reply

A more objective answer is that the lightlike geodesics remain “equally spaced” … but the spacing itself has no observer-independent value!
What does that mean? The cyan curves in this image join neighbouring geodesics with equal-length segments, while always meeting the geodesics at the same angle. But we could always choose a *different* set of curves that made that equal length take a different value.
So in a sense the separation is unchanging, while having no specific value.
Details at https://www.gregegan.net/SCIENCE/PSP/PSP.html#ST
https://media.mathstodon.xyz/media_attachments/files/116/329/543/310/122/188/original/3783458491ea6d9b.mp4

2026-04-01 12:51:23 UTC

- reply

What about lightlike geodesics? Do they converge, diverge, or remain equally spaced? That’s tricky, because in 2D spacetime the only direction perpendicular to a lightlike geodesic is along the geodesic itself, so you can’t measure the separation along a perpendicular vector.
Here lightlike geodesics grow further apart, as measured along the grey circles. This is how the cosmological red shift works: wavefronts in an expanding universe move apart. But that’s a measurement by *particular observers*, at rest in galaxies that are themselves moving apart.
https://media.mathstodon.xyz/media_attachments/files/116/329/537/180/994/757/original/7a479825b590dcca.mp4

2026-04-01 12:50:42 UTC

- reply

But the timelike geodesics behave like lines in the hyperbolic plane: if they start out parallel, they accelerate *away* from each other. The grey curves here, perpendicular to the geodesics and equally spaced along them, have a length that grows with the cosh of the distance.
https://media.mathstodon.xyz/media_attachments/files/116/329/534/499/729/358/original/a6b07e7e0dd5648a.mp4

2026-04-01 12:50:06 UTC

- reply

The spacelike geodesics in de Sitter spacetime behave a lot like great circles on a sphere: if they start out parallel, they accelerate towards each other. The cyan curves here, perpendicular to the geodesics and equally spaced along them, vary in length like the circles of latitude on a sphere crossing the great circles of meridians, with the cosine of the latitude.
https://media.mathstodon.xyz/media_attachments/files/116/329/531/938/495/807/original/c0daf142317255ce.mp4

2026-04-01 12:49:30 UTC

In 3D Euclidean space, the surface equidistant from a point is a sphere, with constant curvature.
In (2+1)-dimensional Minkowski spacetime, the equidistant surface is a (1+1)-dimensional de Sitter spacetime, which also has constant curvature.
The image shows the spacelike (blue), timelike (black) and lightlike (white) geodesics through a point.
https://media.mathstodon.xyz/media_attachments/files/116/329/529/533/704/308/original/d54c4c0caad0c1de.mp4

2026-03-26 11:50:05 UTC

An extremely good article on the horrific bombing of the primary school in Minab that puts it in the context of both the wider history and the recent technological acceleration of the error-prone process of military “targeting”.
https://www.theguardian.com/news/2026/mar/26/ai-got-the-blame-for-the-iran-school-bombing-the-truth-is-far-more-worrying

2026-03-25 12:01:31 UTC

Those staff layoffs and vibe coding seem to be really working out for Amazon ...
https://media.mathstodon.xyz/media_attachments/files/116/289/698/193/994/510/original/b52fef4256748840.png

2026-03-25 10:56:11 UTC

- reply

If you’re wondering how they knew they had 92 antiprotons in their trap, and didn’t lose any along the way, the answer is “image current detection”.
When a charged particle moves past a metal plate, its electric field induces a current in it, with no need for the particle to actual touch anything. That current was monitored to keep track of the contents of the trap.
“Monitoring the antiprotons in real-time: a standard non-destructive image current detection technique is used – this technique is routinely available in precision Penning-trap experiments – so the team gets a continuous signal showing the amount of particles in the trap.”
https://home.cern/science/cern/antimatter-transportation-media-kit

2026-03-25 03:06:14 UTC

- reply

They actually did it!
“On 24 March, a team at CERN transported 92 antiprotons in a specially designed bottle that traps the particles using magnetic fields. The bottle travelled on the back of a truck, for a 30-minute journey ... ”
https://www.nature.com/articles/d41586-026-00950-w

2026-03-19 12:44:03 UTC

The Onion, gloriously on target as ever.
https://theonion.com/the-onions-exclusive-interview-with-sam-altman/

2026-03-18 13:20:17 UTC

- reply

I’m not sure if your “hell no” was directed at me, or at the nightmare Korean chatbot experience, but my own suggestion was *not* for generative AI to be involved at all, but rather for software to be used to make (largely human written) FAQ/help pages easier to access, search and use. This would not be intended to replace human conversations, when necessary, but it would have the potential to address some problems where customers can’t solve some straightforward issues purely because the static FAQs on the company web site are too limited, or it’s too hard to find the relevant page.

2026-03-18 12:46:57 UTC

- reply

No I’m not. Why? Are they already doing something wonderful along the lines I suggested?

2026-03-18 11:43:23 UTC

I have no problem at all — in principle — with automating customer help inquiries. But what this *ought* to consist of, mostly, is exposing more information to the customer in a more accessible, searchable and understandable form.
Carefully written software could supercharge / personalise “help pages” and FAQs on a company’s web site, rendering a significant proportion of help desk phone calls unnecessary. Most of the time when I need to call a company, it’s because their web UI simply isn’t offering me some information that is in their system (and which the help desk staff can see), but that they’ve decided I don’t need to know.
Anyway ... that’s how I think these things *should* work. The link is to reporting on multiple cases where, instead, companies decided to pretend than their “AI” could just mimic human customer service staff and take their place, which of course it couldn’t, and it all went horribly wrong.
[H/T @npub1p20…x3aq ]
https://www.koreatimes.co.kr/southkorea/society/20260316/what-ai-overtaking-jobs-looks-like

2026-03-16 04:01:17 UTC

- reply

I don’t know anything about music in Linux, but this page describes how you can convert AAC files to MP3 or WAV within Apple Music.
https://support.apple.com/en-au/guide/music/musfb0cea9fa/1.5/mac/15.0

2026-03-14 10:15:20 UTC

- reply

It’s puzzling comparing those two accounts, The 2018 Scientific American article talks about a portable trap containing a billion antiprotons, being transported a few hundred metres, though they hadn’t actually done this yet:
“Developing and testing technology for the portable trap will take around four years, with the first measurements scheduled for 2022.”
The 2026 Guardian article talks about 1,000 particles, which they later say are antiprotons. (So apparently not neutral antihydrogen, which would be harder to make and store.) It sounds like the planned drive is quite a bit longer than a few hundred metres in this case: 20 minutes around the CERN campus initially.
So I wonder what happened to the plan to trap and transport a billion antiprotons.

2026-03-14 09:05:58 UTC

Antimatter Trucker is finally a job title! But don’t worry, this isn’t going to be “The Wages of Fear, 2026”:
“The device on Cern’s truck will carry about 1,000 antimatter particles, weighing about a billionth of a trillionth of a gram. Should the containment fail, and the antimatter make contact with normal matter, the resulting pulse of energy would be so feeble, the load doesn’t even warrant a radioactive label.”
https://www.theguardian.com/science/2026/mar/14/please-drive-carefully-scientists-plan-to-transport-volatile-antimatter-for-first-time

2026-03-13 03:36:32 UTC

“AI-powered writing tools are increasingly integrated into our e-mails and phones. Now a new study finds biased AI suggestions can sway users’ beliefs”
“We told people before, and after, to be careful, that the AI is going to be (or was) biased, and nothing helped,” Naaman said. “Their attitudes about the issues still shifted.”
https://www.scientificamerican.com/article/ai-autocomplete-doesnt-just-change-how-you-write-it-changes-how-you-think/

2026-03-12 11:47:59 UTC

- reply

I was a bit stressed before the biopsy, because I thought I remembered reading horror stories about how excruciating this procedure is ... but maybe that’s only much larger extractions for bone marrow transplants, because (with just local anaesthetic) this small biopsy was close to painless.

2026-03-11 05:19:04 UTC

- reply

What’s the public access situation for operating nuclear reactors in Britain? Are there guided tours, or do you stand on a hill outside an electric fence, peering in with binoculars and hoping the security services don’t arrest you?

2026-03-10 12:38:48 UTC

Doctor: I’m not sure what’s going on.
Me: So, time to send Olivia Wilde and Kal Penn to break into my house to search for drugs, toxins and mould, while gossiping about their colleagues’ sex lives?
Doctor: Nah. I think we’ll do a bone marrow biopsy.
Me: It was worth a try.

2026-03-10 05:48:25 UTC

- reply

I didn’t know that about SBS ads! That’s really good for people with gambling problems, and since I’m wholly uninterested in gambling myself, I can choose to have the gambling companies waste their advertising budget on me ... a bit like wiping out a pesky insect by having it try to reproduce with a sterile partner.

2026-03-10 05:31:03 UTC

I’ve yet to watch “Dark Winds”, but since 2 seasons just popped up on Netflix in Australia, a reminder to Australian viewers that it is also available streaming free on SBSOnDemand, who have 3 seasons, and SBS deserve the ad revenue more than Netflix.

2026-03-07 12:37:20 UTC

- reply

Thanks, but no! I actually just rewatched that as well (Herzog loves things going badly with boats), but alas it doesn’t fit.

2026-03-07 12:25:10 UTC

- reply

Thanks, though I’m looking for 20th century movies.

2026-03-07 12:18:12 UTC

Does anyone know what movie [from the 20th century] ends like this?
“A man on a deserted beach had tried to drag a wooden boat into the ocean, to make his escape from some war or revolution. But the boat had been too large, too heavy, and no matter how he strove it had remained firmly beached.”
I had a character in a novel remember that ending ... and now I’ve forgotten the movie I was referencing! I thought it might be Herzog’s “Aguirre, Wrath of God” but I just rewatched that and I was wrong.

2026-03-07 05:06:03 UTC

A new collection of my stories in Korean
잠과 영혼, East Asia Publishing, Seoul, 2026. Translated by Kim Sang-hoon. ISBN13 979-1193078808
Contents:
“Crystal Nights” / “크리스털의 밤”
“This Is Not the Way Home” / “고향으로 돌아가는 길”
“You and Whose Army?” / “너 혼자서?”
“Dream Factory” / “꿈 공장”
“Crisis Actors” / “크라이시스 액터스”
“Mitochondrial Eve” / “미토콘드리아 이브”
“Dark Integers” / “암흑 정수”
“Unstable Orbits in the Space of Lies” / “방랑자의 궤도”
“Sleep and the Soul” / “잠과 영혼”
https://media.mathstodon.xyz/media_attachments/files/116/186/143/998/566/972/original/433c187dbcf103f5.png

2026-03-04 11:03:07 UTC

From gamma decay to photon rockets
Suppose an atomic nucleus with mass m₁ emits a gamma ray and ends up with mass m₂. How fast does the nucleus recoil?
In units where the speed of light is 1, the time and space components of the energy-momentum vectors are:
(𝑚₁, 0) for the initial nucleus
(γ, –γ) for the gamma ray
𝑚₂ (1, 𝑣)/√(1–𝑣²) for the recoiling nucleus
where γ, 𝑣 are unknowns.
Conservation of energy-momentum means:
(𝑚₁, 0) = (γ, –γ) + 𝑚₂ (1, 𝑣)/√(1–𝑣²)
This is solved by:
𝑣 = (𝑚₁² – 𝑚₂²) / (𝑚₁² + 𝑚₂²)
γ = (𝑚₁² – 𝑚₂²) / (2 𝑚₁)
We can also solve for the mass ratio:
𝑚₂/𝑚₁ = √[(1–𝑣)/(1+𝑣)]
But this result also gives us the “rocket equation” for an idealised relativistic rocket, that emits only photons as exhaust and works with perfect efficiency.
For example, it tells us that to accelerate such a rocket to half lightspeed, 𝑣=1/2, it will end up with a payload fraction of √[1/3].

2026-03-02 04:39:09 UTC

- reply

It’s a long time since I wrote that applet, but skimming the source code, there’s a list of disjoint convex polygons into which the original faces are dissected, omitting any that don’t lie on the boundary of the polyhedron as a subset of R^3, and after sorting every such polygon is rendered with no further steps. So, you certainly couldn’t just throw the original faces of a non-convex polyhedron into that list and expect the rendering to work.

2026-03-01 11:52:52 UTC

- reply

They mentioned something at the start then immediately made it clear that what followed would bear no relationship to reality.

2026-03-01 10:38:24 UTC

- reply

But if you’re craving a few more uniform polyhedra, including the self-intersecting ones with star-polygon faces ...
https://www.gregegan.net/APPLETS/26/26.html

2026-03-01 10:24:50 UTC

“The Sticky” is a preposterous, hugely enjoyable Fargo-esque crime caper, with Margo Martindale in overdrive, a cameo by Jamie Lee Curtis, and the Fort Knox of maple syrup.

2026-02-28 02:43:16 UTC

“ChatGPT Health regularly misses the need for medical urgent care and frequently fails to detect suicidal ideation, a study of the AI platform has found, which experts worry could “feasibly lead to unnecessary harm and death”.”
https://www.theguardian.com/technology/2026/feb/26/chatgpt-health-fails-recognise-medical-emergencies

2026-02-27 12:26:26 UTC

RE: https://mastodon.social/@verge/116139531198626533
“Amodei mentioned that “partial autonomous weapons … are vital to the defense of democracy” and that fully autonomous weapons may eventually “prove critical for our national defense,” but that “today, frontier AI systems are simply not reliable enough to power fully autonomous weapons.” He did not rule out Anthropic acquiescing to the military’s use of fully autonomous weapons in the future but mentioned that they were not ready now.”
What fucking deranged fantasy world are these clowns living in. None of this garbage should even be close to anything where lives are at stake.
#note1eu4…ddle

2026-02-25 23:02:55 UTC

- reply

In a recent story of mine, “Death and the Gorgon”, a sheriff’s deputy bonds a little too strongly with his very much non-sentient AI tool and it ... does not go well.

2026-02-25 12:26:40 UTC

- reply

Persian New Year is close.
https://en.wikipedia.org/wiki/Nowruz

2026-02-23 12:35:29 UTC

It’s just crushing to watch an actor of Idris Elba’s calibre deliver a look of utter desolation to the camera, as it dawns on him that the plot of Hijack Season 2 is at least a hundredfold less believable than Season 1.

2026-02-23 11:55:15 UTC

Tomorrow I will be taking a drug to slow down my heart sufficiently for it to be imaged in a CT scan, which seems like the ideal opportunity to also cheat at snooker.

2026-02-22 12:15:44 UTC

RE: https://mastodon.social/@coreyspowell/116114208127759329
NASA on 1 April:
“Our checks have now found leaks of lithium and beryllium!
And boron, carbon, nitrogen, oxygen, fluorine and neon!”
#note1r8l…pdc6

2026-02-20 10:47:04 UTC

Along the world line of a massive particle, we can talk about the “proper time” that would be measured by a clock at rest with respect to that particle.
Along the world line of a photon, the spacetime interval between events is zero. This sometimes causes people who are aware of this from pop science coverage to jump to all kinds of weird conclusions, such as “everything happens all at once for a photon!”
Statements like this are wrong in a number of ways. Firstly, saying “for a photon” is a kind of category error. It is literally impossible to have a clock at rest with respect to a photon, so it is meaningless to talk about proper time here the way you can for a massive particle.
But also, not only are the events along a photon’s world line distinct, you can order them (A, B, C, ...) and you can even say “event C is twice as far along the world line from event A as event B”, notwithstanding zero spacetime intervals all the way.
In special relativity this is simple: just pick any reference frame in which the photon’s world line lies in the (x,t) plane, and use either the x or t coordinates of A, B, C. Different reference frames meeting this condition will assign different values to these x and t coordinates, but they will be connected by linear functions that preserve “C is twice as far from A as B is.”
In general relativity, there is a number called the “affine parameter” along every geodesic, even null geodesics, which can only be changed by a *linear* function if it is to satisfy the geodesic equation. So, like the x or t coordinate in SR, we can use this to order and compare different events on a photon’s world line.

2026-02-20 08:16:49 UTC

- reply

There’s some lighter-hearted satire of this nonsense in (a minor subplot of) “Spare Parts for the Mind” in a recent Asimov’s.

2026-02-20 07:34:51 UTC

A nice summary of the state of the art in indirect detection of cosmic background neutrinos by @npub1hh3…hqxd (https://bigthink.com/starts-with-a-bang/big-bangs-final-prediction/), a useful supplement to a recent discussion on [@dandkuniverse.bsky.social](https://bsky.brid.gy/r/https://bsky.app/profile/dandkuniverse.bsky.social )
In my SF story “Didicosm” the protagonist used indirect neutrino evidence to learn about the precise geometry of the universe:
https://www.gregegan.net/DIDICOSM/Complete/Didicosm.html

2026-02-19 05:29:54 UTC

- reply

Oh, cool! None of the imaging places around me seem to bulk-bill any more.

2026-02-19 04:24:15 UTC

- reply

Did you get it done in a public hospital? My GP referred me to a private cardiologist and I didn’t think to ask if I could go to a public hospital clinic instead. The private cardiologist charges about 3 times the Medicare rebate for everything.

2026-02-19 03:02:31 UTC

Echocardiograms are so expensive that I really think they should offer a 30 percent discount if they fail to locate an alien chest-burster.

2026-02-16 12:02:07 UTC

I know some people are keen on buying their eBooks from Bookshop dot org, so I’m happy to announce that almost* all my self-published DRM-free eBooks are now available there.
* Except for DICHRONAUTS, because the US rights are held by Night Shade so my own edition can only be sold outside North America.
This page lists all the eBooks:
https://www.gregegan.net/BIBLIOGRAPHY/Ebooks.html

2026-02-15 13:31:11 UTC

- reply

Credit card companies make a lot of profit on merchant fees. I don't know how it compares to interest, but it's something like one or two percent on all transactions.

2026-02-15 11:40:56 UTC

“It Was Just An Accident” was powerful, tightly focused on its central quandary but still humanly messy.
Maybe “Death and the Maiden” was first to ask “What if you meet your torturer?” — but sadly the world has more than enough barbarity to justify another version of this story.

2026-02-13 11:50:03 UTC

After listening to an entrepreneur, a philosopher and a psychologist burble confused inanities about “griefbots” for two consecutive nights on the radio [https://www.abc.net.au/listen/programs/the-radio-national-hour/grief-ai-end-death-bot/106338094 and https://www.abc.net.au/listen/programs/the-radio-national-hour/the-promise-and-pitfalls-of-ai-griefbots/106341670] my takeaway was this:
If someone you love needs to hear something from you, for fuck’s sake say it to them now, with your own voice and your own intention, lest some slimy techbro insert himself into their life after you’ve gone, and exploit their loss and pain. Also, tell your spouse/kids/whatever that no, you don’t want your face drawn in crayon on a clumsy digital sock puppet for all eternity.

2026-02-13 06:07:41 UTC

- reply

Sorry for the jargon about “orbits” and D_4, which was just meant to make what I’d done precise for the mathematically inclined.
It’s actually much simpler than I made it sound. @npub1p2n…ayrz created a beautiful set of patterns, and I was curious about how many unique shapes there were among them, as opposed to the same shape appearing in a few different designs, rotated or reflected.

2026-02-13 02:54:48 UTC

- reply

Yes, they are mirror images if you ignore the position of the white space within the square.
So if I was trying to construct, say, a set of unique wooden objects that could be rotated and flipped over, it would only contain one of that pair.

2026-02-12 11:56:26 UTC

- reply

This is very cool!
Here are representatives of the 45 orbits of these figures under the group D_4 (the eight rotations and reflections that are isometries of a square), so no pair of these 45 figures are rotations/reflections of each other.
https://media.mathstodon.xyz/media_attachments/files/116/057/505/145/239/746/original/4831cb4f876ec262.png

2026-02-10 12:18:08 UTC

I very much enjoyed “The Secret Agent”. It reminded me of “Once Upon A Time in Anatolia’’, in the sense it created of being immersed in a very specific time, place and culture, while the flash-forwards expanded the scope and gave everything an eerie poignancy.

2026-02-09 12:47:47 UTC

Both of these Quanta articles describe fascinating mathematical developments — but they also come with encouraging human stories of the ultimate success of persistence over decades by individual mathematicians, fresh collaborations that reignite dormant ideas, and connections to work from long ago.
https://www.quantamagazine.org/long-sought-proof-tames-some-of-maths-unruliest-equations-20260206/
https://www.quantamagazine.org/two-twisty-shapes-resolve-a-centuries-old-topology-puzzle-20260120/

2026-02-09 11:23:38 UTC

- reply

But there is one more easy set of cases.
If y < 1 then L(y) < 0, and if x > y then L(x) > L(y) and x^y > y^x.
That gives us:
(3/2)^(3/4) > (3/4)^(3/2)
but also works when x, y are on different sides of e:
π^(3/4) > (3/4)^π

2026-02-09 11:01:45 UTC

- reply

This instantly gives us, for example:
3^π > π^3
π^4 > 4^π
(9/4)^(3/2) > (3/2)^(9/4)
√3 ^ √2 > √2 ^ √3
But it doesn’t settle cases where x and y are on different sides of e.

2026-02-09 09:56:10 UTC

- reply

Knowing L(x) has a maximum at e also gives us a more general result.
Suppose that either y > x > e or y < x < e, so both x and y lie on the same side of the maximum as each other, and y is further away. We then have:
L(x) > L(y)
log(x)/x > log(y)/y
Multiply through by xy:
y log(x) > x log(y)
Exponentiating both sides:
x^y > y^x

2026-02-08 23:41:04 UTC

For x>0, define L(x)=log(x)/x.
L'(x) = (1-log(x))/x^2 has its sole zero at x=e.
So L(x) ≤ 1/e, with equality only at x=e.
This means:
e log(x) ≤ x
x^e ≤ e^x
for all x>0, with equality only at x=e.

2026-02-08 04:36:42 UTC

Dear MacOS developers
How about you stop uglifying the icons and start disambiguating multiple folders with the same name in the Finder’s Go > Recent Folders menu. Maybe include enough of the end of each path so that every item becomes distinct?

2026-02-06 00:43:28 UTC

- reply

“America is now what anthropologists call a Kardashian Type Three civilisation: more than fifty percent of GDP is in the attention economy.”
https://www.gregegan.net/MISC/RUINS/Ruins.html

2026-02-05 13:47:12 UTC

“the anodic current peak O1 depicts the stoic oxidation of the oxide film, and the enchanting O2 serves as a crescendo harmonizing with the birth of oxygen gas.”
“The cathodic sonnet at C3 serenades the reduction of a wispy oxide film caressing the nickel surface.”
WTF?
“What The Literature Is Filling Up With”
https://www.science.org/content/blog-post/what-literature-filling

2026-02-05 05:40:22 UTC

It’s well known that in special relativity, observers in relative motion can disagree about the shapes and sizes of various objects. But if Alice draws any planar figure, and fires a simultaneous pulse of light from each point on it, perpendicular to the plane, any observer will agree on the figure’s shape and size.
What? Simultaneity is an observer-dependent notion! But if Alice uses *her* definition of simultaneity, she will encode her chosen planar figure in the collection of light pulses in a frame-invariant way.
What happens if Bob, flying past in a spaceship, uses *his* definition of simultaneity to pick out different events on the worldlines of the same light pulses?
Suppose A is a unit-length spacelike vector orthogonal to Alice's worldline, and C is a lightlike vector orthogonal to A. Then the worldlines for two light pulses that Alice considers to be separated by a distance x, one starting at the origin and one starting at x A, will be:
P(s) = s C
Q(u) = x A + u C
where s, u are parameters for how far we go along the respective worldlines.
At s = u = 0 we have what Alice measures:
P(0) = 0
Q(0) = x A
|Q(0)-P(0)|^2
= |x A|^2
= x^2 |A|^2
= x^2
But for *any* s, u:
|Q(u)-P(s)|^2
= |x A + (u-s) C|^2
= x^2 |A|^2 + 2(u-s) A·C + (u-s)^2 |C|^2
= x^2
So whatever events on those two worldlines Bob considers to be simultaneous, he still ends up measuring exactly the same distance between them as Alice does.
https://media.mathstodon.xyz/media_attachments/files/116/016/405/501/533/892/original/797318c1a0d683e1.png

2026-02-04 13:44:59 UTC

I’m sure the writers of “Death By Lightning” know exactly what they’re doing with the anachronistic dialog: endless F-words, “quality time”, “crash and burn” etc. It’s fun, in a West Wing With Beards kind of way. But it still makes me nervous about trusting them on any of the history.

2026-02-03 13:36:11 UTC

- reply

Yes. The first 7 digits are for the manufacturer, which is the same in both cases (they are both “home brand” products of the supermarket, Woolworths).
The one on the left, for the flour, was misclassified as the one on the right, for the raspberries.
It’s all quite strange, but with no hope of looking at their software I doubt it’s possible to diagnose exactly what happened.
https://media.mathstodon.xyz/media_attachments/files/116/006/934/814/911/297/original/150c6fe775ab0ebe.png
https://media.mathstodon.xyz/media_attachments/files/116/006/935/330/825/772/original/72041d50e2bde92b.png

2026-02-03 12:37:06 UTC

- reply

There is a checksum digit that can detect single-digit errors.