Nostr notes by John Carlos Baez

- perhaps because I spend my life talking to and reading the ...

2026-09-20T09:10:39Z

In reply to nevent1q…3d9n
_________________________

- perhaps because I spend my life talking to and reading the writings of mathematicians, I know plenty who consider LLMs unethical for plenty of different reasons. 647 have joined the Association for Human Mathematics:

https://www.ahmath.org/members

and most of these have committed to not using AI in their research. Tasmin Chu has written a list of reasons for not using AI in math:

https://proofsandprompts.com/2026/08/09/the-ai-dissenter-viewpoint/

which curiously does not mention global warming, though it touches on how AI companies are planning to create massive unemployment.

The "skeptics" I was talking about are different: they're people who don't agree that human mathematics is worth preserving, and consider Terry Tao's new initiative a coping mechanism on the part of traditional academics.

We have h₃ = x₁∧x₅∧x₉ + x₂∧x₆∧x₇ + ...

2026-09-19T15:34:19Z

In reply to nevent1q…munm
_________________________

We have

h₃ = x₁∧x₅∧x₉ + x₂∧x₆∧x₇ + x₃∧x₄∧x₈

Compare this to the three parallel lines in the picture! These lines wrap around a lot, so it's a bit confusing, but they go through the points

1,5,9 2,6,7 3,4,8

respectively, and that explains

h₃ = x₁∧x₅∧x₉ + x₂∧x₆∧x₇ + x₃∧x₄∧x₈

So that's the pattern. But what's really going on here? It's rather deep, and so far I've just scratched the surface. So I'll refer you to Elashvili and Vinberg's paper for details. It's connected to the Lie group E8, and other very interesting things.

(7/n, n = 7)

Here's the pattern we're trying to explain: h₁ = ...

2026-09-19T15:19:22Z

In reply to nevent1q…jyfp
_________________________

Here's the pattern we're trying to explain:

h₁ = x₁∧x₂∧x₃ + x₄∧x₅∧x₆ + x₇∧x₈∧x₉
h₂ = x₁∧x₄∧x₇ + x₂∧x₅∧x₈ + x₃∧x₆∧x₉
h₃ = x₁∧x₅∧x₉ + x₂∧x₆∧x₇ + x₃∧x₄∧x₈
h₄ = x₁∧x₆∧x₈ + x₂∧x₄∧x₉ + x₃∧x₅∧x₇

Let's look at

h₁ = x₁∧x₂∧x₃ + x₄∧x₅∧x₆ + x₇∧x₈∧x₉

The red line we saw in our last chart is one of three parallel lines, and these lines go through the points

123, 456, 789

So that's the pattern! Each of the trivectors h₁,h₂,h₃,h₄ corresponds to three parallel lines in our 3×3 grid of points... though sometimes the lines need to 'wrap around'.

Let's compare h₃.

(6/n)

Vinberg and Elashvili classified all kinds of trivectors in ...

2026-09-19T15:09:10Z

In reply to nevent1q…rwmh
_________________________

Vinberg and Elashvili classified all kinds of trivectors in ℂ⁹. But they looked at the space of trivectors modulo SL(n) rather than GL(n). Here SL(n) is the group of n×n matrices with determinant 1. So, they didn't count two trivectors as the same when they differ by an overall scale factor. This actually makes things more pretty, though it means the space has dimension 4 rather than 3.

They showed that up to the action of SL(n), every trivector is a linear combination of these 4 trivectors:

h₁ = x₁∧x₂∧x₃ + x₄∧x₅∧x₆ + x₇∧x₈∧x₉
h₂ = x₁∧x₄∧x₇ + x₂∧x₅∧x₈ + x₃∧x₆∧x₉
h₃ = x₁∧x₅∧x₉ + x₂∧x₆∧x₇ + x₃∧x₄∧x₈
h₄ = x₁∧x₆∧x₈ + x₂∧x₄∧x₉ + x₃∧x₅∧x₇

This looks very scary - almost random! But look at the first term in each sum:

x₁∧x₂∧x₃
x₁∧x₄∧x₇
x₁∧x₅∧x₉
x₁∧x₆∧x₈

Compare it to the chart below. In this chart, I've drawn a 3×3 grid, and all lines through the upper left-hand corner, where the purple line "wraps around". Each line has three points on it! And these points are

1 2 3
1 4 7
1 5 9
1 6 8

matching the pattern we just saw!

But that only explains the first term in each trivector h₁,h₂,h₃,h₄. Can you guess the pattern that explains other two terms in each trivector? I'll show you.

(4/n)

Okay, let's talk about classifying trivectors in different ...

2026-09-19T14:34:58Z

In reply to nevent1q…87z2
_________________________

Okay, let's talk about classifying trivectors in different dimensions.

In n dimensions, the space of trivectors has dimension n choose 3, since you can find a basis of trivectors like xᵢ∧ xⱼ∧xₖ but we have xᵢ∧xⱼ=−xⱼ∧xᵢ .

On the other hand, the group GL(n) of all invertible n×n matrices has dimension n².

The space of 'kinds' of trivectors is the space of trivectors modulo the action of GL(n). Thus, we expect its dimension to be at most n choose 3 minus n². This first becomes positive when n hits 9.

For n less than n, the space of kinds of trivectors is 0-dimensional - in fact a finite set. But when n = 9, it's 3-dimensional, as we might guess from

84 - 81 = 3

(This is not a proof, just a quick guess. But it's correct.)

(4/n)

Also this Aella character: "She’s a data scientist, a ...

2026-09-18T15:54:54Z

In reply to nevent1q…88dr
_________________________

Also this Aella character:

"She’s a data scientist, a sex-research Substacker, and an orgy organizer. And now she is an AI safety advocate, with Plz Don’t Kill Us, an experimental residency program [....] Her concern is that the AI safety movement has a messaging problem." 🤣

https://sfstandard.com/2026/04/22/sex-researcher-aella-hopes-make-ai-doom-go-mainstream/

(3/n)

We know there was an asteroid impact around when the dinosaurs ...

2026-09-13T10:20:27Z

We know there was an asteroid impact around when the dinosaurs went extinct 65 million years ago because we see a rare metal called iridium in the strata from this time - and droplets formed by molten quartz.

But we also see lots of fern spores! This is called a 'fern spike', and it's not the only one. Ferns spread rapidly using spores, and were the first to recolonize the land after the Cretaceous-Paleogene and Triassic-Jurassic extinction events.

I like ferns. Knowing this makes me like them even more.

Thanks to Nicole Parsons (npub174q…ckk8) for telling me about fern spikes. Read more here: Nicole Parsons (npub174q…ckk8)/117201753504722650">https://mathstodon.xyz/Nicole Parsons (npub174q…ckk8)/117201753504722650

Nostr event nevent1qqs87tft5j8paxdwn40al7tuw2jr8am9j5yrlh8dgfemxmkd7gv8jgszyzdx5x5walgt20t7pjtx4v6mmyzp2yjx7qa3h6v05rfdpm45jspjq6fzeas

2026-09-10T15:54:37Z

In reply to nevent1q…9wxz
_________________________

- [@gro-tsen.bsky.social](https://bsky.brid.gy/r/https://bsky.app/profile/gro-tsen.bsky.social ) is very active on Bluesky. I can tell Gowers is too busy to waste time on social media, but he occasionally posts a note on Bluesky, and it's worthwhile. He never posts anything here anymore. Daniel Litt has been weighing a bit on the AI business.

In general I find the math scene on social media extremely depressing these days. If I post anything deep and technical I get almost no replies. Of course that's an odd thing to do on social media, but I used to get enough responses to make it feel worthwhile. Now I find it more rewarding to talk to a LLM.

Realizing more and more how modern-day AI taps into some very old ...

2026-07-06T08:38:32Z

Realizing more and more how modern-day AI taps into some very old urges.

So many would like "unsurpassed mastery without any effort", and would summon a demon to get it.

The publisher Springer claims it can only discuss why it ...

2026-06-28T10:35:36Z

The publisher Springer claims it can only discuss why it retracted some of Planck's papers with... Planck!

But until they hold a seance, they are happy to sell you completely redacted versions: blank pages for $39.95 or more each.

Make a binary tree listing all matrix products of 1 0 and 1 1 1 1 ...

2026-05-28T12:59:43Z

Make a binary tree listing all matrix products of

1 0 and 1 1
1 1 0 1

as below. If you multiply

1
1

on the left by all these matrices, you get the Calkin-Wilf tree. If you multiply

1 1

on the right by these matrices, you get the Stern-Brocot tree! These are both individually famous, but Katherine Stange's article explains them together:

https://math.colorado.edu/~kstange/papers/CWSB.pdf

- The usefulness of AI is not dependent on a precise definition ...

2026-05-27T11:49:51Z

In reply to nevent1q…s4g0
_________________________

- The usefulness of AI is not dependent on a precise definition of intelligence any more than the usefulness of a plow is dependent on a precise definition of "furrow". It's very hard to precisely define almost anything, except in the realm of math and maybe physics, but we limp along nonetheless.

For example, mathematicians are now cracking unsolved problems using LLMs, and this sort of thing will continue regardless of whether LLMs are 'truly intelligent" or not. See below for the latest big surprise.

I disagree with Penrose's criticisms, since they're based on bad physics (assuming quantum entanglement is important in microtubules) and bad logic (assuming that while computers are limited by Godel's theorem, we somehow are not). He's a great physicist, but he's wrong here.

https://interestingengineering.com/ai-robotics/openai-paul-erdos-geometry-problem-cracked

"Every city in the Midwest has a half-finished canal in ...

2026-05-26T07:38:43Z

"Every city in the Midwest has a half-finished canal in it."

I love canals, so this makes me want to do a tour of midwest US cities and its half-finished canals - and a boat tour of its *finished* canals. But I'm in Scotland so... let me do it online.

First up: Indianapolis!

(1/n)

[image from Sue is Writing Solarpunk 🌞🌱 (npub1f5n…2tcs)/116635972190246192">https://mathstodon.xyz/Sue is Writing Solarpunk 🌞🌱 (npub1f5n…2tcs)/116635972190246192]

Similarly, a 'diminished seventh' is any 4 notes that ...

2026-05-14T18:01:13Z

In reply to nevent1q…nv59
_________________________

Similarly, a 'diminished seventh' is any 4 notes that form a square!

Here's what an augmented triad sounds like:

https://en.wikipedia.org/wiki/Augmented_triad

and here's a diminished seventh:

https://en.wikipedia.org/wiki/Diminished_seventh_chord

(2/2)

An 'augmented triad' is a kind of chord with 3 notes. If ...

2026-05-14T17:54:47Z

An 'augmented triad' is a kind of chord with 3 notes. If you draw the 12 notes on the piano as equally spaced dots on a circle, an augmented triad is simply any chord that forms an equilateral triangle!

(1/2)

In clean samples of blue bronze, the phasons can travel faster ...

2026-05-11T12:31:44Z

In reply to nevent1q…pg6p
_________________________

In clean samples of blue bronze, the phasons can travel faster than sound - indeed, this is the most famous example of a supersonic lattice excitation. It's possible because the restoring force for the sliding mode of charge density waves has nothing to do with the ordinary stiffness of the crystal.

So: confine some electrons to one-dimensional chains, cool the result down, and they organize themselves into a rippling wave whose collective motions can outrun sound!

But as we cool blue bronze still further, the density waves lock into step with the underlying atomic lattice, and all these effects go away.

(6/n, n = 6)

In blue bronze the Peierls instability kicks in below about 180 ...

2026-05-11T12:24:35Z

In reply to nevent1q…fmgg
_________________________

In blue bronze the Peierls instability kicks in below about 180 K. The octahedra in the chains bunch into pairs, the crystal abruptly stops conducting, and a charge density wave forms along each chain of octahedra.

Once the charge density wave has formed, it has two natural ways to wiggle:

• its amplitude (how strongly bunched the ripple is) can pulse;
• its phase (where exactly the peaks of the ripple sit along the chain) can shift.

Pulsing the amplitude costs real energy. But sliding the phase - pushing the entire density wave bodily along the chains - costs essentially nothing. Why? Because nothing in the crystal cares where the peaks of the ripple happen to land. This is a strange gift of the fact that the charge density waves have a wavelength that's an irrational multiple of the spacing between atom pairs. No position is preferred over any other, so the wave can slip freely!

The resulting slow, almost-free sliding excitation is called a phason, and very gentle, long-stretching versions of it cost vanishingly little energy to excite. There's a deep principle at work here, known as Goldstone's theorem, that says whenever a system spontaneously settles into one of infinitely many equivalent configurations, it must come with a corresponding gentle shimmer mode that explores the alternatives.

So the Peierls instability in blue bronze doesn't just give a static charge density wave: we get phasons too!

(6/n)

Here's a longer version: Blue bronze is molybdenum oxide ...

2026-05-11T12:13:20Z

In reply to nevent1q…2lch
_________________________

Here's a longer version:

Blue bronze is molybdenum oxide K₀.₃MoO₃ — a deep-blue, metallic-looking crystal in the family of alkali molybdenum bronzes first prepared by Wold, Kunnmann, Arnott and Ferretti in 1964. As far as we can tell, it doesn't exist in nature. The name "bronze" is jargon inherited from the sodium tungsten bronzes Wöhler made back in 1825. It refers to the brassy luster of these compounds, not to the copper-tin alloy.

Blue bronze is built from parallel chains of molybdenum and oxygen atoms threading through the crystal, with potassium ions tucked between the layers. The electrons free to carry current are essentially trapped on these chains, like cars on a one-lane highways with no exits. This makes blue bronze a textbook example of a quasi-one-dimensional metal.

This is the setup for a remarkable result Rudolf Peierls established in the 1930s: a one-dimensional metal cannot stay metallic in its ground state! Instead, its atoms spontaneously bunch themselves into pairs, raising a barrier that stops electrons from flowing freely and turning the metal into an insulator. Simultaneously, there is a slight periodic rippling in the density of conduction electrons — a charge density wave. But the wavelength of the charge density wave can be an irrational multiple of the spacing between atom pairs! And in blue bronze, this does happen. It's thus an example of an 'aperiodic crystal'.

(5/n)

By the way, blue bronzes are just one class of ...

2026-05-11T11:46:01Z

In reply to nevent1q…gfyy
_________________________

By the way, blue bronzes are just one class of molybenum-oxygen-alkali metal compounds. Boost the amount of alkali a little and you get 'red bronzes'. Boost it a lot more and you get 'purple bronzes'. All these look and act metallic, hence the name.

https://en.wikipedia.org/wiki/Molybdenum_bronze

In blue bronze, the wavelength of these charge density ripples is ...

2026-05-11T11:39:57Z

In reply to nevent1q…5tk5
_________________________

In blue bronze, the wavelength of these charge density ripples is not a rational multiple of the distance between octahedra! As a result [insert math here] it's easy for these ripples to slide back and forth along the octahedra chains.

So we get 'phasons'!

(3/n)

https://en.wikipedia.org/wiki/Phason

'Blue bronze' is a crystal that can vibrate in two ways: ...

2026-05-11T11:19:18Z

'Blue bronze' is a crystal that can vibrate in two ways: ordinary sound waves, and charge density waves that move faster than sound. Thanks to quantum mechanics, waves act like particles. So in blue bronze we get 'phonons' - particles of sound - but also 'phasons' moving faster than sound!

(1/n)

- since it's four-dimensional it's called a polychoron ...

2026-05-06T12:05:46Z

In reply to nevent1q…la7p
_________________________

- since it's four-dimensional it's called a polychoron rather than a polyhedron, and some call it the hexacosichoron.

Jason Hise (@npub1mcu…g2al) thinks he's found a way to ...

2026-05-06T10:27:16Z

Jason Hise (Entropy (npub1mcu…g2al)) thinks he's found a way to partition the tetrahedral faces of the 600-cell into 6 isomorphic sheets of 100 tetrahedra! Each sheet is organized like a 10 × 10 lattice on a Clifford torus. Here is his animation:

https://entropygames.net/files/shared/600_cell_six_sheets.html

He asks: has anyone seen this before?

In case like me you hadn't been keeping up on the Ukraine ...

2026-04-13T20:19:28Z

In reply to nevent1q…we9m
_________________________

In case like me you hadn't been keeping up on the Ukraine war, you might like to watch Anders Puck analyze the current situation and prognosticate here.

He and the host agree that Ukraine is now killing off Russian soldiers faster than Russia can recruit them. I never know if I can trust these numbers, but Ukraine claims to have killed ~35,000 last month and has a plan to kill 50,000/month - mainly by increasing its production and use of drones, I guess. Anders Puck and the host say that Putin will be driven to conscript troops, which so far he's avoided doing - and that this will force him to increase various repressive measures. But they think this will only slow the decline in Russian power: the only way out for Putin will be to stop the EU from supporting Ukraine financially. The defeat of Orbán makes that harder for him. They say he'll eventually threaten the Baltic states with war. He's already threatened Estonia, Finland and Latvia for letting Ukrainian drones fly over their territory.

In short, they think Putin is losing but he'll become more dangerous as he becomes more desperate.

(2/2)

https://www.youtube.com/watch?v=iqkWf12Wcxo

Wow! While I was driving to New Mexico to visit archaeological ...

2026-04-13T20:05:21Z

Wow! While I was driving to New Mexico to visit archaeological sites, Ukraine was hammering oil refineries in Russia with drones! How come none of you told me? 😜

Urkaine sent drones over 1000 kilometers to destroy oil storage tanks and shipping infrastructure at the ports of Ust-Luga and Primorsk on the Gulf of Finland. These ports were rendered unable to ship any cargo. Reuters estimated that Russia lost about 40% percent of its oil export capacity while Bloomberg put the figure at 43%

As you know, Putin had been enjoying the oil price spike due to the Iran war. But this changes things. He'll try to rebuild the ports and/or reroute the oil. But until he gets an effective defense against these drones, the Ukrainians can just repeat what they've done.

(1/2)

https://www.aljazeera.com/features/2026/4/6/smell-of-war-comes-to-st-petersburg-as-ukraine-hammers-russian-refineries

- This is what you're missing: the bigger a black hole, the ...

2025-12-15T10:40:11Z

In reply to nevent1q…yt9e
_________________________

- This is what you're missing: the bigger a black hole, the less "dense" it is, if by density we mean its mass divided by the cube of the radius of its event horizon.

That's because the radius of a black hole is proportional to its mass!

(All this is for nonrotating black holes, and for nonrotating black holes a better definition of "density" might be the mass divided by 4π/3 times the cube of its radius. But that doesn't affect my point; black holes in nature don't generally rotate so much as to deviate strongly from being spherical, and the Universe doesn't seem to be rotating much at all, though sometimes experiments seem to show otherwise.)

This chart shows all objects in the Universe, arranged by mass ...

2025-12-14T19:13:54Z

This chart shows all objects in the Universe, arranged by mass (vertical) and radius (horizontal).

The edge of the upper left corner is the "Schwarzschild radius" - anything here becomes a black hole, so we don't expect there's stuff in that corner.

The edge of the lower left corner is the "Compton wavelength" - anything here has a size so small that measuring its position that accurately would require enough energy to create a new one.

These two corners intersect in a white dot misleadingly called "instanton" - that's not the right name for it. It's really a black hole so small its Schwarzschild radius is its Compton wavelength. Its mass is the Planck mass, and its radius is the Planck length.

The pink strips of slope 3 are lines of constant density. For example "QGP" is the density of quark-gluon plasma, "BBN" is the density of the universe when Big Bang nucleosynthesis was going on, and so on.

The Earth is only slightly more dense than a flea, and that makes sense!

The black dot labeled "Hubble radius" is the whole observable universe.

I like this chart a lot but I would get rid of that word "instanton". It's from here:

C. H. Lineweaver and V. M. Patel, “All objects and some questions”, American Journal of Physics 91 (2023), 819-825. Free at https://www.researchgate.net/publication/374231292_All_Objects_and_Some_Questions

See the alt text for more!

wrote: "how are the three 8-dimensions irreps of Spin(8) ...

2025-12-05T17:08:23Z

In reply to nevent1q…y7yk
_________________________

wrote: "how are the three 8-dimensions irreps of Spin(8) related to each other and to octonion multiplication?"

The most beautiful approach is to *start* with the three 8-dimensional irreps V, S₊, S₋ of Spin(8) and the god-given map

t: V ⊗ S₊ → S₋

that you have on general principles. These irreps each have an inner product and thus a norm preserved by Spin(8), and in 8 dimensions we have something unusual:

||t(v \otimes s)|| = ||v|| ||s||

Then we can do this. Pick any unit vector v in V. Then t(v ⊗ -) is an isometry from S₊ to S₋. Then pick a unit vector s in 𝑆₊. Then t(- ⊗ s) is an isometry from V to S₋. Using these isometries we can identify V, S₊ and S₋ as vector spaces (not as reps). Call this vector space 𝕆. This identifies v, s, and t(v ⊗ s) with each other, and call this vector 1. We now have the octonions with their multiplication t(- ⊗ -) and unit 1.

But maybe you know all this and are wanting to start with the octonions and make it into an irrep of Spin(8) in 3 ways. That's what Yokota's Theorem 1.14.2 accomplishes:

https://golem.ph.utexas.edu/category/2025/09/a_shadow_of_triality.html#c069068

More precisely the *proof* of the theorem shows how to get, for any element of Spin(8), two others. This implements the outer automorphisms of Spin(8). From these and the usual 'vector' representation of Spin(8) (really SO(8)) on 𝕆 you get the other two 8-dimensional irreps.

- Sobex is right. When you have a lot of atoms and a really good ...

2025-04-08T22:34:08Z

In reply to nevent1q…jnwh
_________________________

- Sobex is right. When you have a lot of atoms and a really good detector, you can detect a few decays per year and determine the half-life of that kind of atom even if it's very long. The longest-lived unstable isotope known, tellurium-128, has a half-life of 2.2 × 10²⁴ years!

https://en.wikipedia.org/wiki/Isotopes_of_tellurium

Some atomic nuclei are round. Others are football-shaped. Some ...

2025-04-08T16:59:20Z

Some atomic nuclei are round. Others are football-shaped. Some are even stranger. Some are pear-shaped! Some have a 'halo' of protons and neutrons orbiting a smaller core. They're all just minimizing energy - unless they're in a temporary 'excited state'. But the forces between protons and neutrons are so complex that this can make lots of things happen.

Just as noble gases are exceptionally inert because they have a filled shell of electrons, the roundest and most stable nuclei are those with a filled shell of protons - and, separately, a filled shell of neutrons. Filled shells happen when we hit a 'magic number'. The magic numbers are

2, 8, 20, 28, 50, 82, and 126.

I want to show you some cool patterns in these numbers. But first, some examples:

• Hydrogen, with just one proton, is perfectly round, but that's an exception.

• Helium-4 has 2 protons and 2 neutrons, and that's the second most common element in the universe.

• Oxygen-16 has 8 protons and 8 neutrons, and this is the most common form of oxygen.

• Calcium-40 has 20 protons and 20 neutrons, and this is the most common form of calcium. It's also the heaviest stable nucleus that has the same number of protons and neutrons! Heavier nuclei need more neutrons to be stable, but calcium-40 is stabilized by the fact that 20 is a magic number.

• Calcium-48 has 20 protons and 28 neutrons. It's not stable, because it has too many neutrons - but it has a half-life of 64 quintillion years, so it's damned close to being stable, thanks to the power of magic numbers.

• Lead-208 has 82 protons and 126 neutrons. It's the most common form of lead.

(1/2)

I didn't know that wolves can pollinate plants. This is an ...

2024-11-29T18:16:07Z

I didn't know that *wolves* can pollinate plants.

This is an Ethiopian wolf, the most endangered predator in Africa. These wolves like to lick sweet nectar off a flower called the Ethiopian red hot poker. And as they go from flower to flower, their muzzle gets dusted with pollen! So while it's not yet *proved* that these wolves are important as pollinators, it seems awfully likely.

Scientists think these wolves don't get much of their caloric intake from this nectar. The “dessert hypothesis” says the wolves just like the taste!

For more, check out this open-access paper:

• Sandra Lai, Don-Jean Léandri-Breton, Adrien Lesaffre, Abdi Samune, Jorgelina Marino and Claudio Sillero-Zubiri, Canids as pollinators? Nectar foraging by Ethiopian wolves may contribute to the pollination of Kniphofia foliosa, https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecy.4470.

When I was a kid, under the influence of science fiction, I ...

2024-11-10T18:57:46Z

When I was a kid, under the influence of science fiction, I dreamt I was going to the moon. It was so exciting! When I woke up I was crushed. When I grew up, I realized sending humans into space is a waste of money now, though it was good to prove we could do it. If we're serious about exploring airless environments with lots of radiation, we should be developing cyborgs and better AI.

But meanwhile we can explore space from the comfort of home using space probes and telescopes. We just need to get better at explaining to ordinary folks how cool this is. I can get thrilled by graphs and tables of data, because I can vividly imagine what they mean. Most folks would probably prefer a good animated video.

MeerKAT is an amazing array of 64 radio telescopes in South Africa. They want to expand this to the Square Kilometer Array, which will actually consist of thousands of telescopes in South Africa and Australia. But it's already seeing great stuff. Like just this June they found a weird thing that flashes like a pulsar - but really slowly, once every 54 minutes instead of a thousand times a second like an ordinary pulsar.

That's insane! What is this thing? We don't know, and that's exactly why it's cool. But for some guesses, read on.

(1/2)