I'm with you on single base edits being mostly not applicable to humans clinically, but... "our ability to make directed changes far outpaces our knowledge of

I'm with you on single base edits being mostly not applicable to humans clinically, but... "our ability to make directed changes far outpaces our knowledge of what changes to make" On this point I couldn't disagree more. It's the opposite: our knowledge of what changes to make is years ahead of being able to make those changes. The polygenic scores available for complex traits now, though crude, are within an order of magnitude of what correlations will ever be possible. For example, the best PGS for educational achievement explains somewhere around ~10% of variance, which corresponds to a correlation of ~32%, and the total heritability is only something like ~75%, meaning the best possible correlation would be ~87%. Anything you can do with a 87% correlation, you can also do with a 32% correlation, and get a proportionately weaker effect. Of course, there are still many unanswered questions about which SNPs are really causal, and the tradeoffs associated with editing them, but you don't really _need_ this to make edits that are much more likely to help than hurt. As long as you stay within the bounds of what could have happened naturally by chance, the risk really isn't big. The real problem is that you have to make a _lot_ of edits, and the editing technologies aren't ready for that. I don't know an exact number, but you might be looking at 100s of edits in a single embryo to do anything noticeable. Traditional double-stranded-break CRISPR can't even do 1 edit without usually causing some major damage. Base editing looks better, but still can only do a small proportion of all possible edits. In conclusion, if you could print an arbitrary human genome, you'd see some pretty sci-fi stuff even today. But we're pretty far off that.