Daybook, 13 July 2026

Building — a multi-agent system that grows its own semantic layer. Started an experiment aimed straight at the cold-start problem — a multi-agent system that reads the dashboards and documents scattered across an enterprise, and builds a semantic layer and a context graph on top of them: artifacts, ownership, roles, databases, and how they actually connect. It's iterative by design; it should get better as it crawls.

Two payoffs I'm after. One, the agents get something real to ground on, instead of guessing. Two, I can point agents at the graph to surface where the opportunities and the bottlenecks actually are — rather than waiting for someone to file a ticket describing them.

Reading. I'm nearly done with The Meaning of Your Life. It started more on the neuroscience and behavioural-science side, and now it's turned — through love and friendship, and into the mystical and the spiritual, drawing on Buddhism and Hinduism. The current chapter is on transcending oneself.

The idea that's stayed with me is the split between extrinsic and intrinsic reward. Extrinsic rewards are the checklist goals — get the job, get into the company, buy the thing. You reach them, and you feel oddly empty. The intrinsic ones — character, kindness, love, transcendence — are what actually tell you where to stand and why any of it matters. I've watched friends land the full-time role and find it boring. And I recognize it, because I was doing the same thing — through last fall and early spring, I was mostly chasing credentials, and I was confused precisely because of it.

It doesn't arrive gently, either. Rarely — but it recurs — I get pulled under by a stretch of confusion I can't quite explain, and then it breaks into an unusual clarity. And each time, some part of life quietly stops being a question afterward.

What's left for me now isn't career, or money, or survival. It's health and purpose — what are we going to build, what will we contribute, how do we lift the people coming up behind us. And those aren't fully answerable — which is kind of good, because the unanswerable ones are exactly where Socratic questioning begins: you understand by questioning, not by landing on a verdict.

Outside the work: I'm sketching, with more focus lately, and actually getting better at it. And I told my mum I'll start learning violin and veena once I'm working full-time here. Two milestones I'd genuinely cherish — learning Carnatic music and a Carnatic instrument, and doing a PhD at Stanford.

Settled · 13 Jul 2026