Lucas Beveridge is talking about something he and Sam Piper did for Atomos Nuclear and Space. Suppose we only shipped one reactor to a HEO. That reactor hands over energy to tungsten plates - one per mission. The plate then delivers the kilowatts. Here's how it was tested; here's what was tested.
At HEO - as in, GEO or higher - you can fuel up a lot of space tuggery. Take Tory Bruno's Blue Ring, which needs one tonne of xenon for four of payload. That reactor looks like it would in fact deliver the 5.5 kW needed to propel to Gateway's lunar NRHO. To get a mission rolling "for weeks or months", Beveridge prescribes Scandium-46 with its low 84-hour halflife. Maybe tugs; maybe Venus injection. Sc-46 is allowed to deorbit after three years.
The test used Manganese-56 which decays in 2.6 hours. This can deorbit rather sooner. Maybe use the Mn-56 for high-thrust tugs between HEOs and be done with it. Like if you just want to hit the Spaceline cable at the 80 Mm this blog has promised.
Naturally none of this is to run on Earth (the least radioactive Co-60 is legendary). The idea is that the reactor - which is way too high up for Earthlings to care - creates neutrons, which then irradiates the neutral cobalt or scandium. The hot stuff, in turn, rams gamma (and not more neutrons) into the metal plate, which then heats up. Or maybe just heats up some beryllium, which I'm seeing recommended for high-enthalpy rockets for Martian landers.
For longer missions, he's talking Cobalt-60 at its 5.27 year halflife. Beveridge is suggesting this for a Uranus orbiter, or series of orbiters. First it has to get there but we can probably do that with a Starship. And/or with a fusion-driven ship launching off Deimos.