This.
The biggest failure of education today is that its oriented towards an outcome (tests / uni) that is dubious at best, detrimental at worst
OP is correct to identify dev cost as holding the key to software-based education. Here is why:
Education today is the remnant of the industrial assembly line model. That worked out great...for fridges and cars.
Children are not fridges. Children are not cars.
This education where 1-size-fits all produces the worst outcomes. It holds back brilliant students and leaves the challenging cases behind.
Software will solve all that due to (eventual) low cost customization.
In order for this to work, many planets have to align. In particular, some idealists have to let go of the model where all students should learn x topic. The entire curricula must be on the table
In other words, society must come to accept that
1) learning will not be uniform ie some students will get more out of school than others and
2) students will pass on topics we take for granted. Biology , civics, algebra etc....all gone, provided that students can explore AND dive deep into an alternative topics they care about, whether that is mechanics, geology, philosophy etc.
In short. Kids today are going to school memorizing some info, yet learning nothing. Thet should instead be able come out masters of topics they are interested in, if indeed that's all they care about.
This would be prohibitively expensive to do with teachers as it would require a 1:1 ratio.
Software can guide that journey.