Okay, now the physics. Because his thermos analogy self-destructs on closer inspection.
He's right that vacuum eliminates conduction and convection. But his conclusion - that radiated heat is "ineffective at processor temperatures" - is exactly backwards.
Stefan-Boltzmann law: P = εσAT⁴
Radiation scales with the fourth power of absolute temperature. At 80°C / 353K, a high-emissivity surface radiates on the order of 800-900 W per square meter. Raise the temperature and heat rejection climbs fast. Vacuum doesn't kill radiation - in space, radiation is the mechanism you design around.
That's where the thermos analogy falls apart. A thermos suppresses all three heat-transfer paths: vacuum handles conduction and convection, while silvered low-emissivity walls suppress radiation. Space radiators do the exact opposite - maximize area, emissivity, view to cold space, and heat transport to the radiator panels. He accidentally described the problem, then missed the standard engineering answer.
Reports on the AI1 concept describe roughly 110 m² of deployable liquid radiators for a compute payload in the ~120 kW sustained / 150 kW peak range. That's not an ignored problem. That's the cooling system.
Could the design still have hard thermal-engineering challenges? Absolutely - sunlight, Earth infrared, albedo, pointing, radiator temperature, pump loops, mass, and margins all matter. But "space is a vacuum, therefore it dies in minutes" isn't the argument. Satellites, probes, the ISS, and spacecraft electronics have been rejecting waste heat by radiation for decades.