I think Nate Silver either could not or chose not to evolve with the field of data analysis. He gained prominence in a time period when data science was in it's infancy and by data science I mean the application of AI to data analysis which had been the industry standard before then.
Now, there are free Python libraries that can automatically produce thousands and thousands of models like Nate made, test them all, rank them, add remove/remove all sorts of variables, automatically test which regression models best fit the data (poisson, LSS, etc) all automatically.
Plus why wouldn't anyone just chose to be a pundit given that choice? Punditry is one of the least accountable, most upwardly failing jobs you can get. If he's known for predicting elections otherwise, you can get those wrong and look uncredible in your primary goal. As a pundit you might get a bunch of "this you" tweets when you're wrong bfd
people don't look to pundits for right and wrong. they look to pundits for reassurance.