James
@astrocrash.net
Data scientist, gadget enthusiast, star destroyer, San Diego lover.
Been tackling OEIS sequences the past several days. There are ~400k sequences now, so finding the ones to go after is tricky. So I built a graph visualization of all of the nodes/relations using Cosmos. Really impressive software, ~60fps on a massively complicated graph.
September 28, 2026 at 4:43 PM
Been tackling OEIS sequences the past several days. There are ~400k sequences now, so finding the ones to go after is tricky. So I built a graph visualization of all of the nodes/relations using Cosmos. Really impressive software, ~60fps on a massively complicated graph.
Another project : my daughter asked a question, "what color was the sky when the dinosaurs were around?" So...I computed it as accurately as I could! I have a writeup with the details coming soon, but in the meantime, explore what Earth's sky looked like throughout its life.
earthsky.astrocrash.net
earthsky.astrocrash.net
September 26, 2026 at 11:14 PM
Another project : my daughter asked a question, "what color was the sky when the dinosaurs were around?" So...I computed it as accurately as I could! I have a writeup with the details coming soon, but in the meantime, explore what Earth's sky looked like throughout its life.
earthsky.astrocrash.net
earthsky.astrocrash.net
Gave my personal website a glow-up! Pulled all the projects I've been involved in. Wanted it to look unique. :-) astrocrash.net
September 26, 2026 at 9:04 PM
Gave my personal website a glow-up! Pulled all the projects I've been involved in. Wanted it to look unique. :-) astrocrash.net
A very minor math breakthrough teaser: 38456, 2490634, 550112272
September 26, 2026 at 3:03 AM
A very minor math breakthrough teaser: 38456, 2490634, 550112272
starseti.astrocrash.net
I redid the search of Hippke/Learned from 10+ years ago for *all* variable stars from ZTF, ASAS-SN, TESS, and Kepler. I also redid the physics; my conclusion is that Cepheids (the orig. Cepheid Internet paper) are not feasible, but RR-Lyrae stars are. Civs need to be ~KII.
I redid the search of Hippke/Learned from 10+ years ago for *all* variable stars from ZTF, ASAS-SN, TESS, and Kepler. I also redid the physics; my conclusion is that Cepheids (the orig. Cepheid Internet paper) are not feasible, but RR-Lyrae stars are. Civs need to be ~KII.
Star SETI — modulation candidates
starseti.astrocrash.net
September 23, 2026 at 4:56 PM
starseti.astrocrash.net
I redid the search of Hippke/Learned from 10+ years ago for *all* variable stars from ZTF, ASAS-SN, TESS, and Kepler. I also redid the physics; my conclusion is that Cepheids (the orig. Cepheid Internet paper) are not feasible, but RR-Lyrae stars are. Civs need to be ~KII.
I redid the search of Hippke/Learned from 10+ years ago for *all* variable stars from ZTF, ASAS-SN, TESS, and Kepler. I also redid the physics; my conclusion is that Cepheids (the orig. Cepheid Internet paper) are not feasible, but RR-Lyrae stars are. Civs need to be ~KII.
well, i tried: searching 100,000 variable stars, no signs of aliens.
still figuring out the best way to share these results since i a) don't have time to write a paper unassisted but b) do think the results/method should probably be shared. will prolly go with an ai-forward platform, not the arxiv.
still figuring out the best way to share these results since i a) don't have time to write a paper unassisted but b) do think the results/method should probably be shared. will prolly go with an ai-forward platform, not the arxiv.
September 23, 2026 at 1:28 AM
well, i tried: searching 100,000 variable stars, no signs of aliens.
still figuring out the best way to share these results since i a) don't have time to write a paper unassisted but b) do think the results/method should probably be shared. will prolly go with an ai-forward platform, not the arxiv.
still figuring out the best way to share these results since i a) don't have time to write a paper unassisted but b) do think the results/method should probably be shared. will prolly go with an ai-forward platform, not the arxiv.
If 2 mathematicians with AI and Claude Max almost solved a millennium problem, that’s a pretty good investment…hybrid human/AI is where it’s at.
This phase where we’re uncomfortable with the notion of working with AI to accelerate advancement will pass, I think.
This phase where we’re uncomfortable with the notion of working with AI to accelerate advancement will pass, I think.
September 19, 2026 at 5:10 AM
If 2 mathematicians with AI and Claude Max almost solved a millennium problem, that’s a pretty good investment…hybrid human/AI is where it’s at.
This phase where we’re uncomfortable with the notion of working with AI to accelerate advancement will pass, I think.
This phase where we’re uncomfortable with the notion of working with AI to accelerate advancement will pass, I think.
I'm convinced that progress on mathematically adjacent scientific publications will be *dominated* by AI, starting roughly now with the release of Fable 5.1. This includes astronomy, physics, etc. It runs absolute circles in speed and accuracy around any human.
September 3, 2026 at 10:08 PM
I'm convinced that progress on mathematically adjacent scientific publications will be *dominated* by AI, starting roughly now with the release of Fable 5.1. This includes astronomy, physics, etc. It runs absolute circles in speed and accuracy around any human.
…is it normal for ArXiv to post its papers over an hour late nowadays?
September 1, 2026 at 1:04 AM
…is it normal for ArXiv to post its papers over an hour late nowadays?
Just couldn't help myself. Two more papers on the way, soon. :-)
August 30, 2026 at 6:48 PM
Just couldn't help myself. Two more papers on the way, soon. :-)
After building the math sniper, suggesting improvements for other people’s papers, I realized I could do the same for myself: a draft I always wanted to finish, but never did. Well, over the course of a few nights chatting with Claude in great detail…I finally finished it! 8 years in the making.
August 28, 2026 at 7:44 AM
After building the math sniper, suggesting improvements for other people’s papers, I realized I could do the same for myself: a draft I always wanted to finish, but never did. Well, over the course of a few nights chatting with Claude in great detail…I finally finished it! 8 years in the making.
Working with AI, I just proved a theorem! A little one, but it's pretty cool to be able to do this now. Found a paper using the sniper app I posted earlier that could be potentially improved; GPT Terra had an initial attempt that got partway there, then I finished it off with Fable.
August 23, 2026 at 6:45 AM
Working with AI, I just proved a theorem! A little one, but it's pretty cool to be able to do this now. Found a paper using the sniper app I posted earlier that could be potentially improved; GPT Terra had an initial attempt that got partway there, then I finished it off with Fable.
stealth drop of a big idea: sniper.astrocrash.net
Math Sniper
sniper.astrocrash.net
August 22, 2026 at 10:36 PM
stealth drop of a big idea: sniper.astrocrash.net
I’ve now seen 10 schools “regress” and start requiring the Physics GRE again. I maintain the Google sheet that tracks this. My question is: with the rise of LLMs making it much easier to “fake” other components of an application, is the test now seen as the lesser evil? Is there a point to tracking?
August 12, 2026 at 4:30 PM
I’ve now seen 10 schools “regress” and start requiring the Physics GRE again. I maintain the Google sheet that tracks this. My question is: with the rise of LLMs making it much easier to “fake” other components of an application, is the test now seen as the lesser evil? Is there a point to tracking?
Impulse buy for a very special event I hope happens in the next few years
July 10, 2026 at 2:50 AM
Impulse buy for a very special event I hope happens in the next few years
A paper I wrote over a decade ago, "the fastest unbound stars in the universe", should be close to testable with JWST or Rubin now. Would be cool if some astronomers out there took a look at their data for stars in the middle of nowhere moving at 10% the speed of light :-) arxiv.org/abs/1411.5022
The Fastest Unbound Stars in the Universe
The discovery of hypervelocity stars (HVS) leaving our galaxy with speeds of nearly $10^{3}$ km s$^{-1}$ has provided strong evidence towards the existence of a massive compact object at the galaxy's ...
arxiv.org
June 1, 2026 at 5:41 PM
A paper I wrote over a decade ago, "the fastest unbound stars in the universe", should be close to testable with JWST or Rubin now. Would be cool if some astronomers out there took a look at their data for stars in the middle of nowhere moving at 10% the speed of light :-) arxiv.org/abs/1411.5022
Took a look at Karpathy's autoresearch repo and man it's kinda...shit
April 9, 2026 at 6:23 PM
Took a look at Karpathy's autoresearch repo and man it's kinda...shit
Uh anyone else seeing this massive sell order on oil futures placed in the last few hours? TACO insider trades?
April 7, 2026 at 10:34 PM
Uh anyone else seeing this massive sell order on oil futures placed in the last few hours? TACO insider trades?
“Genocide isn’t the answer Mr. President”
- Some NYT opinion wank
- Some NYT opinion wank
April 7, 2026 at 9:59 PM
“Genocide isn’t the answer Mr. President”
- Some NYT opinion wank
- Some NYT opinion wank
Reposted by James
If there were ever a time to invoke the 25th amendment it would be now. Donald Trump has threatened Iran with genocide. This has to end!
credits: @nowthisimpact
credits: @nowthisimpact
April 7, 2026 at 9:11 PM
If there were ever a time to invoke the 25th amendment it would be now. Donald Trump has threatened Iran with genocide. This has to end!
credits: @nowthisimpact
credits: @nowthisimpact
People shouldn’t dismiss this as impossible or stupid, there’s a lot of dumb money out there and it has a good chance of happening. If it happens astronomy from the ground is going to be nearly impossible. The sky is going to constantly look like a swarm of buzzing bees to the naked eye.
SpaceX render comparing Starship with what it calles "AI Sat Mini', which I estimate to be 174 metres across the solar arrays
March 22, 2026 at 2:50 PM
People shouldn’t dismiss this as impossible or stupid, there’s a lot of dumb money out there and it has a good chance of happening. If it happens astronomy from the ground is going to be nearly impossible. The sky is going to constantly look like a swarm of buzzing bees to the naked eye.
I think people maybe should start looking up more, even during the day, to understand just how much extra crap is in orbit now. That white dot is a satellite, visible in broad daylight at 1pm.
March 19, 2026 at 10:44 PM
I think people maybe should start looking up more, even during the day, to understand just how much extra crap is in orbit now. That white dot is a satellite, visible in broad daylight at 1pm.