Most of modern climate science is still taking real-world measurements, the modelling is based on that and is intended to refine not replace real world evidence.
The modelling is also intended to examine how the climate responds to various forcings and scenarios, and has been on many occasions confirmed once the data were collected.
And that shouldn't be surprising. General Circulation Models work from first principles, so it should be expected that modelling experiments would predict behaviour. And they have.
Coldstream routinely says the models haven't predicted anything, which is pure ignorant bunkum. Models have predicted all sorts of things we now know to be true, from the temperature profile of the layers in our atmosphere, to the ENSO associated water vapour feedback, the expansion of Hadley cells, even the increase in coastal upwelling of water.
Even Arrhenius' simple model in 1896 predicted that nights would be warming more than day, that the polar regions would warm faster owing to polar amplification, that the Arctic would warm faster than Antarctica, and that the winter would warm faster than the summer.