Hacker News Reader

And then an 18-to-20-something-year training run is required for each individual instance.

I know right, such a waste. Plus it's so random on how they will turn out!

Any suggestions on how to reduce that waste?