It's a little difficult to summarize what has happened since, since
so much has happened. This is a golden age of human genetics which has no precedent, and major new findings are being uploaded to Biorxiv at an almost daily rate.
I think the short version would be that the BGI project made a bad bet on a particular genetic architecture of intelligence (rare variants with large boosting effects) which would have implied that a tiny sample of whole genomes from very smart people would yield a lot of genetic variants; this was wrong and they turned up little, and then the project was further damaged by BGI's disastrous strategic choice to try to develop its own genome sequencer tech rather than using Illumina's, which set it back by years. What actually wound up happening in 2013 was that a big consortium of American & European research groups, the SSGAC, pooled simple SNP data (ie the sort of thing you get from 23andMe) rather than whole genomes, for 100k people, and got the first genetic variants for intelligence (rendering the BGI project largely moot). SSGAC has kept expanding its sample size, and the massive UK Biobank dataset started coming online (just completed their full n=500k release!), and the combination has yielded hit after hit. The last paper from last year turned up 172 SNP hits, and there will be another paper this year which will expand that even more.
In conjunction with this, the IQ polygenic scores have been used for all sorts of exciting things: people have been computing genetic correlations with health and psychiatric traits, been looking at cross-national differences, examined changes over thousands of years using ancient DNA, decreases over the past century confirming dysgenics, been deploying methods for detecting selection in different populations, and nailing down the architecture as having the same heritability in normal people as in very intelligent people with no special rare variants detected, and being mostly additive with most of that coming from common SNPs but the rest from relatively rare harmful variants suggesting a mutation load explanation of why there is any genetic contribution.
As far as embryo selection goes, '5-15 points' is very much an upper bound assuming the entire population, great polygenic scores (better than the ones in 2017, perhaps more like 2020-level), and ignoring the practicalities like loss of embryos in the IVF process. I've tried to work through a full analysis of what embryo selection could do, and it's not earth-shaking unless there are huge improvements: https://www.gwern.net/Embryo%20selection (Which is why I'm more interested in iterated embryo selection and genome synthesis now.)