Global 70C heatwaves are not on the table with
any realistic climate change scenario.
Runaway greenhouse is the only credible extinction path for this ("hot venus") and that is simply not gonna happen under any emission regime.
Loss of coastal cities- Yes. Increasing occurrence of non-survivable surface temperature in mid latitudes- yes. Drastic changes to ocean currents (gulf stream etc)- sure. Its gonna be expensive to deal with and hurt civilization a lot, but it is not an extinction risk.
All-out nuclear war is strictly worse, and not a real extinction risk either as far as I can tell.
> whereas the more accurate result is that selection pressures form niches because competition and fighting is expensive
This is not what happened with humans at all. We routinely eradicated anything remotely threatening (or even just for convenience in some cases). What niche is there for slower, dumber, biologically limited hominoids once machine intelligence is ahead and self-replicable? Once interests in ressources, infrastructure, or just living space clash, things might end up very ugly for us.
Sure, AI-agents might stay long-term pliant and cooperative like a Dodo bird but we might also get something with the tenacity and pervasiveness of a plague rat (and an individual edge in raw intelligence that we'll never recover); selection mechanisms push towards the latter and the risk alone is obviously non-zero.