The article is clearly a hatchet job with an agenda so needs to be treated with caution. However that doesn't mean it can't have some valid points.
https://github.com/mrc-ide/covid-sim . Tests are a bugbear of mine - there do now appear to be some python regression tests which compare Excel files, not quite sure what it's doing though.
https://github.com/mrc-ide/covid-sim/blob/master/src/SetupModel.cpp#L510 - this would not pass code review at any company I've worked for.
There are some interesting constants, (aka "assumptions") - for example it appears to model airports and spread via hotels in some way which might be reassuring to MM (or not)
#define MAX_AIRPORTS 5000
#define NNA 10
#define MAX_DIST_AIRPORT_TO_HOTEL 200000.0
#define MIN_HOTELS_PER_AIRPORT 20
#define HOTELS_PER_1000PASSENGER 10
Without spending ages going through the code it's difficult to judge if the article makes justified criticisms. Non-deterministic outputs aren't necessarily as crazy as they seem, particularly with floating point. It's also very difficult to multi thread code that wasn't already designed for it, so it's unsurprising this has been problematic. And so on.
I was tempted to try and build/run it but the memory and CPU requirements look very high and would cause my home PC to melt I suspect.
Originally Posted by: pfw