He's not entirely wrong. Essentially, the more evidence you find of the Earth being more than 4000 years old, the more evidence you have against a non-deceiving god having created it 4000 years ago. If there's a 0.1% chance that a god will erase all evidence of his existence, then we can only get 20 bits of evidence against him.

The problem is most likely that he's overestimating the probability of a god being deceitful (conjunction fallacy), and that he's forgetting that it's equally impossible to find evidence for such a god (conservation of expected evidence).