For years I've trusted the database to just work. mini_db is me deciding to build one instead — parser first, then storage, indexing, and a query executor — in C++, deliberately climbing under the abstractions I've been shipping on top of.
It's slow going, and that's the point. This is the fundamentals layer that turns 'engineer who ships' into 'engineer who understands the machine.' It's the same instinct pulling me deeper into AI: I don't want to only call the API; I want to understand what's underneath it.
Fundamentals and fluency compound
The fastest way I've found to understand a system you depend on is to try to rebuild a small version of it — you learn exactly where the hard parts hide. And the two tracks reinforce each other: shipping fluency makes the fundamentals concrete, and the fundamentals make everything I ship sharper.