It's also that if you take things that improve one side of mental performance it's likely to harm another. This isn't massively surprising to me: you'd expect that if upping a single hormone level or whatever would simply improve performance overall then evolution would have 'found' it. But presumably the same is true of giving performance-enhancing drugs to less intelligent animals - or, for that matter, giving people steroids etc. to increase their physical performance.

But just because drugs to make you run faster might lower your life expectancy, that doesn't mean our current running speed is the best evolution or technology can achieve. The problem is that any complex adaptation, like intelligence, is going to be a 'sweet spot' in the sense that a random massive change in a single factor will make it less succesful. That doesn't mean that evolution, or potentially much more sophisticated technological enhancement, can't improve matters.

Also, the 'something's going to get worse' principle only holds if what we consider bad is the same as what evolution selects against. It could in principle be true that humans became much more intelligent if they lost something that made them capable of defending themselves, reproducing, making allies or whatever. If our aims are different to what benefits our genes' survival, we may well be able to improve on nature: as we do with artificial sweetners, sex with condoms and other cunning tricks.

Of course, that assumes that autism should be considered a mental disorder. Many of those on the autism spectrum don't, whereas most of those with depression or high levels of anxiety do consider their condition to be a disorder. It looks a lot to me like status quo bias: if being more intelligent will cause our minds to become qualitatively different then we shouldn't try to be more intelligent.