Meaning it captures most of the primary effects but is still very simplistic. It’s good for highlighting general trends, but not good at making good fidelity estimates. As noted by others, he makes a lot of simplifying assumptions. And then he makes the same mistake early models did, he extrapolates from a low fidelity model. Basically, he hand waves about R (which defines spread) assumes that he is dealing with a homogenous population (which were not), doesn't bring in any external data about what has happened in other countries, and wraps it up with a declarative statement that we therefore need to continue to social distance, which he neither supports nor proves with his data. (Note that I'm assuming he is not lumping everything we're doing to decrease R into "social distancing". If he does include all our current efforts, I still take issue with the idea that we have to continue all the current efforts, as he hasn't proved that all steps are necessary)