What makes timsort fast is how it adapts to common patterns in the data (sorted runs, etc). Paul isn't trying to do any of that.