Andrew Quinn replaced a 3 GB SQLite database with a 7 MB FST (finite state transducer) binary, and buried in a footnote is a sharper observation about how engineers actually learn.
Quinn's argument: you need to reinvent roughly 4 to 5 wheels in most domains, closer to 20 to 30 in rigorous fields like mathematics or computer science, to reach the real frontier of a problem space. Not zero wheels, not a thousand. The guilt of rediscovering known solutions is a trap. Each deliberate reinvention, paired with directed questions, moves you faster than passive study or reading documentation ever will.
The FST compression result is the headline, but the footnote is the reason to click through. Quinn is describing a specific, countable heuristic for skill development, not a vague growth mindset platitude. That precision is worth reading in context.
[READ ORIGINAL →]