Models are more accurate than before, but the next reports based on them will have a too high warming bias. So people need to see beyond that known flaw.

Not sure how, when there are zero credible plans for cutting emissions (they are accellerating), or sequester 40+ Gigaton CO2 per year. Nothing so far addresses extra releases from burning biomass, acidification of oceans and declining albedo.

What do you think of people feeding climate models into economic models? Would any errors get amplified?

For sure, but they may also use simple projection or just current levels already, as negative feedbacks are already accounted for. These will start breaking from 2030 however, so we might see further acceleration until equalization is reached.

Any serious measure can use uncertainty and other measures in combination.

Economy tends to undershoot existential risks, so think bailouts until the supply-lines break down. We already saw the world flirting with this just from the Corona-situation.

Mitigations tend to be reactive, rather than proactive, and may overshoot for individual random events compared to overall preparedness.

Such complex models will come with flaws. The concern is however that scientists find few significant redeeming factors overall. The concern is that the older models are too conservative even.

Extinction is wildly unlikely. Humans occupy a wide range of climates, from very cold to very hot, and even the most extreme predictions don't make the earth completely unlivable.

Complete collapse of the current civilization and population drops of 90% or even 99% aren't impossible. But there's an enormous difference between a 99% loss and a 100% loss.