⚡️🌙
@dystopiabreaker.xyz
recovering cryptographer building ML models, doing systems work, security, etc.
when you understand he was leading meta ai their recent (lack of) meaningful frontier output makes more sense
January 23, 2026 at 12:51 AM
when you understand he was leading meta ai their recent (lack of) meaningful frontier output makes more sense
this mf is NOT actually swimming!!
January 23, 2026 at 12:49 AM
this mf is NOT actually swimming!!
space literally does “autocool” everything, point a panel at deep space and that’s what happens. this is how every cooling system on every spacecraft works and part of why spacecraft have low-emissivity MLI insulation everywhere to manage this. it’s just slow at most human temperatures
January 22, 2026 at 3:01 AM
space literally does “autocool” everything, point a panel at deep space and that’s what happens. this is how every cooling system on every spacecraft works and part of why spacecraft have low-emissivity MLI insulation everywhere to manage this. it’s just slow at most human temperatures
…yes? that’s why i was showing stefan-boltzmann calculations and computing the required rad sizes for various T?
January 22, 2026 at 2:46 AM
…yes? that’s why i was showing stefan-boltzmann calculations and computing the required rad sizes for various T?
again, you can use convection or conduction _on the spacecraft_ to move heat around, into a radiator that… radiates, not conducts or convects, it out into space
January 22, 2026 at 2:43 AM
again, you can use convection or conduction _on the spacecraft_ to move heat around, into a radiator that… radiates, not conducts or convects, it out into space
it depends what you mean; you can have convective or conductive heat transfer on spacecraft, loop cooling, and so on, but ultimate _rejection_ is only through blackbody radiation
January 22, 2026 at 2:30 AM
it depends what you mean; you can have convective or conductive heat transfer on spacecraft, loop cooling, and so on, but ultimate _rejection_ is only through blackbody radiation
a few watts of that is just blackbody radiation, though
January 22, 2026 at 2:26 AM
a few watts of that is just blackbody radiation, though
everything is always emitting heat under all 3 transfer methods: conduction, convection, and radiation. in that example the dominant effect is probably convection followed by conduction
January 22, 2026 at 2:22 AM
everything is always emitting heat under all 3 transfer methods: conduction, convection, and radiation. in that example the dominant effect is probably convection followed by conduction
you’d need thousands, and probably concentrated around SSO, but space is huge and you can steer them
January 22, 2026 at 2:19 AM
you’d need thousands, and probably concentrated around SSO, but space is huge and you can steer them
heat isn’t really transmitted “onto” anything, the energy just leaves the object with the photon. you can think of whatever is in front of the radiator as a sink, if you point at deep space you’re pointing towards the 2.7K background, at the sun and you’re having a bad time
January 22, 2026 at 2:07 AM
heat isn’t really transmitted “onto” anything, the energy just leaves the object with the photon. you can think of whatever is in front of the radiator as a sink, if you point at deep space you’re pointing towards the 2.7K background, at the sun and you’re having a bad time
>99.9% of the sun’s heat reaches us through electromagnetic radiation (photons), not particle radiation. a radiator panel emits exactly the same type of radiation, just all infrared because it’s cooler. all objects emit photons according to their temperature, this is blackbody radiation.
January 22, 2026 at 2:04 AM
>99.9% of the sun’s heat reaches us through electromagnetic radiation (photons), not particle radiation. a radiator panel emits exactly the same type of radiation, just all infrared because it’s cooler. all objects emit photons according to their temperature, this is blackbody radiation.
the main thing is that they are assuming:
- you literally have a datacenter-sized satellite, instead of a bunch of modules
- that you run your radiator at room temperature?? instead of junction temperature or higher
- you literally have a datacenter-sized satellite, instead of a bunch of modules
- that you run your radiator at room temperature?? instead of junction temperature or higher
January 21, 2026 at 9:57 PM
the main thing is that they are assuming:
- you literally have a datacenter-sized satellite, instead of a bunch of modules
- that you run your radiator at room temperature?? instead of junction temperature or higher
- you literally have a datacenter-sized satellite, instead of a bunch of modules
- that you run your radiator at room temperature?? instead of junction temperature or higher
yeah but we already have examples of B200 scale heat rejection in space, we can do that, we don't have examples of 'b200 that survives space' or 'infiniband-scale interconnect over sat-to-sat laser'
January 21, 2026 at 9:32 PM
yeah but we already have examples of B200 scale heat rejection in space, we can do that, we don't have examples of 'b200 that survives space' or 'infiniband-scale interconnect over sat-to-sat laser'
we already have space datacenters (satellite ISPs), and the reason we did that is universal global availability + no local permission required every time you want to expand the infrastructure. similar thing here.
January 21, 2026 at 9:28 PM
imo the heat rejection engineering challenges are really not hard compared to the challenge of 'make a petaflop-scale compute module that survives space' or 'have the launch capacity to launch these' or 'have the interconnect to make them work'
January 21, 2026 at 9:27 PM
imo the heat rejection engineering challenges are really not hard compared to the challenge of 'make a petaflop-scale compute module that survives space' or 'have the launch capacity to launch these' or 'have the interconnect to make them work'
that's because the ISS's EATCS runs at human-habitable temperatures, and the T^4 scaling of stefan-boltzmann is brutal. if you raise the temperature to even just ~80C (B200 comfortable junction temperature), you halve this, and you don't need as much margin for non-manned spacecraft
January 21, 2026 at 9:12 PM
and then even after you've done that you have to design high bandwidth networking, like a massive mesh net of node-to-node laser or something
January 21, 2026 at 9:07 PM
and then even after you've done that you have to design high bandwidth networking, like a massive mesh net of node-to-node laser or something
yeah i mean my assumption is that launch costs will keep coming down but the much harder problem is actually designing a tensor compute unit that survives space
January 21, 2026 at 9:05 PM
yeah i mean my assumption is that launch costs will keep coming down but the much harder problem is actually designing a tensor compute unit that survives space
which part is unclear? the spacecraft is a black body radiator, which in every textbook is referred to as doing heat transfer, without anything to transfer through and with only the CMB to transfer "to"
January 21, 2026 at 8:57 PM
which part is unclear? the spacecraft is a black body radiator, which in every textbook is referred to as doing heat transfer, without anything to transfer through and with only the CMB to transfer "to"
you transfer heat to the cosmic microwave background which is 2.7 Kelvin and there is no luminiferous aether
January 21, 2026 at 8:44 PM
you transfer heat to the cosmic microwave background which is 2.7 Kelvin and there is no luminiferous aether
the total rejection is almost certainly in the 10s of megawatts if you sum everything in LEO
January 21, 2026 at 8:31 PM
the total rejection is almost certainly in the 10s of megawatts if you sum everything in LEO
...also i mean we already have swarms of satellites rejecting 8xB200 scale heat all the time
January 21, 2026 at 8:30 PM
...also i mean we already have swarms of satellites rejecting 8xB200 scale heat all the time
you literally can just say 'we'll radiate better' because that's how black body radiation works if you even just modestly increase the temperature
January 21, 2026 at 8:23 PM
you literally can just say 'we'll radiate better' because that's how black body radiation works if you even just modestly increase the temperature
also im pretty sure the ISS's own ammonia coolant is pressurized at like 200+ psi
January 21, 2026 at 8:22 PM
also im pretty sure the ISS's own ammonia coolant is pressurized at like 200+ psi
water? glycol? liquid metal? but this is moving the goalpost lol
January 21, 2026 at 8:21 PM
water? glycol? liquid metal? but this is moving the goalpost lol