Actually, that specific denial is limited to two items: gender based discrimination by Tom Preston-Werner and his wife, and the existence of a discriminatory/hostile workspace. Not coincidentally, these are the two claims which would have been against the law, and also the least of the claims brought forth by Horvath (though it's the angle latched onto by the media) - she instead claimed personal harassment from Horvath's wife, and the non-responsiveness of HR to these issues.

The only real gender based problems she had were related to male co-worker's responses to female co-workers hula-hooping, as I remember it.

There wouldn't be much evidence of personal harassment, particularly the kind described by Horvath.

"[D]irect and testable assertion" sounds scientific, certainly. It would be interesting to hear how you would flesh that out, though. For example, when they say "no evidence exists," don't we immediately run into the question of whether the claims of Ms. Horvath don't actually constitute evidence? What does constitutes "evidence"? Is that question testable? If Ms. Horvath's claims don't count as evidence, then why does Github's claim that "no evidence exists" count? What test or tests do you propose?

Criminal and civil investigations are not scientific investigations. They may make use of science, but they are not scientific. To import the language of science--actually neo-Popperian pop-cultural language about science--into a discussion of an internal investigation into harassment claims is highly suspect.

Your claim aside, these are weasel words. The way to discover the fact that they are is to appropriately apply language akin to that of jurisprudence, rather than inappropriately apply that of science.

I'd like to see how you (or anyone here) propose to test it. This isn't science; it's a corporate press release.

Do we know the name of the "full, independent, third-party" investigator(s)? Do we know who paid for them, their methodology, and whether they reported to Horvath, giving her oversight?

Or are investigators the same, like software contractors?

You and I or 'tptacek probably cannot test it, but Horvath's lawyer may be able to sue to test it.

It would be very dumb of GitHub to state that there is no evidence if there is evidence that Horvath could turn up. (Granted, sometimes companies are dumb.)

No, the blog post says: "The investigation found no evidence".

Failure to find is distinct from a negative result.

You can't test for failure to find, just as you cannot test for statistical insignificance.