Riyaz and research
What years of Hindustani classical practice taught me about systems: improvisation inside constraints, repetition without boredom, and staying with complexity.
I started Hindustani classical training around the age of five, long before I wrote a line of code. People treat that as a charming biographical footnote. It is closer to the operating system everything else runs on.
Repetition without boredom
Riyaz is repetition. You sing the same phrase, the same raag, the same patterns, for years. The skill is not surviving the repetition — it is staying awake inside it, noticing the small thing that is slightly off on the four-hundredth pass.
That is the exact muscle good systems work demands. Reading a latency graph for the tenth time and catching the one spike that does not belong is the same act of attention as catching a note that has drifted a few cents flat.
Improvisation inside constraints
A raag is not freedom. It is a strict set of rules about which notes are allowed, in which order, with which emphasis. The art is improvising inside that structure — finding something new without breaking the form.
Replace "raag" with "production system" and the sentence still holds. Good engineering is improvisation inside constraints: latency SLOs, hardware budgets, the model you actually have. The constraint is not the enemy of the work. It is the thing that makes the work mean something.
Staying with complexity
The longest-running gift of the practice is patience for complexity that does not resolve quickly. A difficult composition does not yield in an afternoon. Neither does a real research question about what inference costs under load.
Discipline under uncertainty.
Both halves of my life turn out to be the same instruction, written in two different scripts. Before AI, there was riyaz — and the riyaz is still running.