The best thing we can do to see if the models work is to look how they performed compared to reality in the decades since we have climate models. Turns out: They worked really, really well.
https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1029/201...
Right, but again this goes back to OP's point that there is a survivorship bias in the models which tends the community to accept the models that both "predict" the past as well as reinforce the narrative that temperatures will continue to increase.
Having said that, the reality remains that, even without the models, the CO2 levels are increasing, they are well correlated to temperature increases, and they are expected to increase further - so VERY PROBABLY - climate change is a very serious massive problem. This is all true even without the models.
When it comes to the reliability of models, the devil is in the details. ...but, of course, no one should assume that a model turning out to be wrong or biased suddenly negates all the other evidence we have. Models are just one piece of a probability decision tree we need to navigate.
The main article's set of models are different than the ones referred to in another link someone in this thread provided, and I suspect the article you've linked will also use yet another unique set.
If it was me trying to persuade people, and all historic climate change models did indeed work really, really well, I'd make an authoritative web page that lists all of these models including their inputs and outputs, and show a detailed quantitative comparison of their outputs. I would think this approach would be more persuasive in multiple ways than hundreds of separate articles written in a narrative style, lacking numbers and consistency in sets of models.
Having said that, the reality remains that, even without the models, the CO2 levels are increasing, they are well correlated to temperature increases, and they are expected to increase further - so VERY PROBABLY - climate change is a very serious massive problem. This is all true even without the models.
When it comes to the reliability of models, the devil is in the details. ...but, of course, no one should assume that a model turning out to be wrong or biased suddenly negates all the other evidence we have. Models are just one piece of a probability decision tree we need to navigate.