Could another factor impeding broad spectrum research be that the two big broad spectrum approaches in use today, chemotherapy and radiation, are both pretty brutal to the patient and many people believe they'll be looked back on as barbaric in the future? That's not to say that future broad spectrum treatments necessarily have to share these characteristics, but there's sort of a "guilt by association" problem.
That still leaves the question of how the reader is to distinguish a sound (speech) from a description of sounds.
Run-on refers primarily to topic drift, not word count. You can have a fairly short run-on sentence where the end of the sentence is talking about something different than the beginning. Frequent use of "and" between clauses is a clue.
I don't think children have any more difficulty learning to speak English than other languages. The difficulty comes in learning to spell in writing and, to a lesser extent, learning to pronounce written words when writing. Btw, there's actually much more regularity in English spelling/pronunciation than may appear, and than is routinely taught. Much of the "weirdness" is the result of historical processes which are fairly regular in themselves, once you know the rules.
Given that the default, non-quotation text is not, in general, describing sounds in the environment, why do you think a reader would interpret unquoted text as environmental sounds rather than as simply more of the author's description of goings on in the scene? I can see that presenting spoken words in some format that allows or encourages their interpretation as environmental might be artistically useful, I just don't see that removing the quotation marks from otherwise-quoted dialog accomplishes that.
Does it seem likely that a trait that has survival significance (in a highly social animal such as a human) would be emergent? Even if it might have been initially, you'd think selective pressure would have brought forth a set of genes that have significant influence on it.
There are many orders of magnitude between hundreds and millions. Also, perch are near the top of the food chain, while these mirror life forms would be near the bottom.
Putting aside for the moment the fact that even "intelligence" is hardly a well-defined and easily quantified property, isn't it rather a giant leap to say we even know what a "better" personality is? I might agree that some disorders are reasonably well defined, and those might be candidates for trying to "fix", but if you're trying to match greater intelligence with "better" personality I think you first need a far better notion of what "better" personality actually means.
I think the "second" question about second order effects was really the main question here. If the intentional beneficial effects don't quite add, that's no great tragedy, but if combining multiple edits produces unexpected changes, some of which are bad, that's kind of a deal killer. I don't find the chart you reference to be very convincing, since it only lists a handful of characteristics that are sufficiently common to have names and have been studied in such an analysis. For every one of those there are likely countless less frequent and/or more subtle "bad" variations whose correlation with the things we're trying to fix we have no idea of. Informally, don't you have to wonder why, if a small number of edits would seem to lead to a clearly superior genome, natural selection hasn't happened upon some of those combinations already? How can we know what combinations were "tried" earlier in our evolution with long-term negative consequences?
While you could give your internal AI wide indiscriminate access, it seems neither necessary nor wise to do so. It seems likely you could get at least 80% of the potential benefit via no more than 20% of the access breadth. I would want my AI to tell me when it thinks it could help me more with greater access so that I can decide whether the requested additional access is reasonable.