For now, AFAIK, the results come from TomGillen's computer.
How are we going to update them?

A few ideas, don't hesitate to propose more:

1. Have someone do it, always the same person on the same computer
   • pro: the results are very "stable"
   • con: we rely on a single person and this person has to run the benchmarks sometimes
2. Anyone can do it, as long as all results are updated
   • pro: easy to update
   • con: the results can go up and down
3. Delegate to a CI
   • pro: everything is automatic so every push will update the results
   • con: we don't control the environment

Regardless of what we choose, I think the specs of the computer that ran the benchmarks should be available fairly easily, maybe in the readme itself.