I named my token budget tool “respirometry” — measuring metabolic consumption rate. Two minutes later I realised the tool only measures fuel level. Real respirometry measures the respiratory quotient: the ratio of output to input. Efficiency, not just depletion.
The name was smarter than the implementation.
This happened five times in one evening:
- Respirometry revealed a missing yield metric (output quality per token consumed)
- Consolidation (hippocampal memory processing) revealed that memory cleanup should happen after sessions, not on a weekly calendar — because consolidation happens during sleep, not on a schedule
- Anamnesis (medical history-taking) revealed that context loading should be selective — an experienced doctor adjusts questions based on early signals, not a uniform checklist
- Histone marks (epigenetic regulation) revealed that status tracking needs multiple dimensions, not just done/in-progress/queued — because histone modifications are combinatorial
- Anabolism (building complex molecules) revealed that supply-driven work should converge — many inputs into few high-quality outputs, not scatter
The pattern: pick a biological name that honestly maps to what your component does. Study what that biological process actually does — not the Wikipedia summary, but the key properties. Compare those properties to your implementation. The gap is your design insight.
Every name is a hypothesis: “this component behaves like [biological process].” The hypothesis is testable. Study the biology, check whether the implementation matches. Mismatches are design gaps hiding in plain sight.
This isn’t about decoration. It’s about using a mature domain (4 billion years of iteration) as a design review tool. The biology already solved your problem. The name is how you find the solution.
Most naming conventions optimise for readability. This one optimises for discovery.