tl;dr:"The practice of breaking up a large change into smaller, individually reviewable PRs which can depend on each other, forming a DAG." Ben discusses advantages, including that stacked PRs are easier & quicker to review, since there are fewer changes to sign-off on, and easier to rollback if they cause breakage.
tl;dr:"In other words, Difficult Problems involve operating in a higher dimensional space than Hard Work. Strategies that can collapse the higher dimensional Difficult Problem into a lower dimensional form of Hard Work reduce the cognitive costs to solving the problem." Ben discusses his approach to balancing the two.
tl;dr:Ben's biggest takeaway is that there's "a significant difference between a data structure as an abstract mathematical concept (“a B⁺-Tree”) and concrete implementations (“SQLite’s database format”)." The optimizations to implementations "won’t improve the BigO characteristics of a data structure, but will have significant “real world” implications on performance & usability of a DB."