Agreed. And more; the Macbooks are pretty much the same - some are god approximations, some are terrible, all of them are recognisably a MacBook. And if you start using it they can train on it.

The problem isn't the test, its that is a public test.

Simon has previously said he has a list of secret prompts (at least one of which he "burned" as a demonstration a while ago). That's what makes it a good test - his commentary on the public test is something of a proxy for non-public tests. This makes it a good benchmark.