Do you happen to have 2 months old predictions of the models you find non-trivial? My sense is that this is weather vs. climate prediction. They may be able to do a somewhat reasonable job a few days ahead, but beyond that it's all about common sense, i.e. similar to the historical record, as limited as it may be. They probably constantly retrofit the models with current data, which makes it look like the models where accurate all along.

Common sense can also work in the short term. E.g. Austria (and Netherlands) seem to have crested the cases wave [1], which I predicted a 3 days ago by looking at the death rates plateauing, and noticing that the same pattern happened in US southern states and BG/RO [2]. I wonder what said models were saying 3 days ago, if such models exist for AT/NL and if they happen to publish their historical predictions.

[1] https://news.ycombinator.com/item?id=29334288

[2] https://ourworldindata.org/explorers/coronavirus-data-explor...