01
How to benchmark Python without lying to yourself
Design repeatable timing experiments that measure the workload you actually care about.
Independent field notes / Vol. 01
Precise explanations of the machinery, patterns, and tradeoffs that make Python feel deceptively simple.
Read the cover storyThe index
4 notes, edited for signal.
Design repeatable timing experiments that measure the workload you actually care about.
Build a small import hook while learning how finders, module specs, loaders, caches, packages, and failure cleanup divide responsibility.
Use cProfile and pstats to locate expensive call paths before choosing what to optimize.
Use disassembly as versioned evidence, connect stack operations to source semantics, and avoid turning CPython opcodes into false contracts.
Editorial policy
No listicles. No folklore. No just trust me abstractions.
Every note should leave you with a better model of the system, not just another snippet to paste.