{

}

My now page as of Feb 15th '26

It's inspired by Derek Sivers'nownownow  project, and is meant to capture a point-in-time snapshot of what I was upto at the time. I've written about how I built and maintain ithere.

Listening

Music: Loner - Barry Can't Swim

Reading

Book: Flesh - David Szalay

Watching

TV: Industry - Season 2
Movie: One Battle After Another

Weeknotes

  • Speed-read my way through Flesh by David Szalay. The plot (and I use the word loosely) felt thin, which made it harder to explain why I liked the novel as much as I did. I even watched Dua Lipa’s interview with Szalay about the book. It somehow made me appreciate Dua Lipa more (if that’s even possible), yet left me none the wiser about why I liked the damn book!
  • After groundlessly dismissing the efficacy of knee sleeves to help with my knee pain during tennis, I’m now a convert. Had a surprisingly pain-free few sessions after trying them for the first time last week. This week I’ve also added an elbow support strap to my protective paraphernalia. I’m now only a few steps away from showing up in full medieval knight cosplay for tennis games.
  • After agonising over it until the last minute, I finally decided against traveling to Mumbai for the John Mayer gig. Disappointed, but I’m taking solace in earmarking the saved money for a solo trip I’m planning in March - destination TBD.
  • Did a round of load-testing for AutoPCR’s Training Agent this week. It’s slightly absurd to stress-test a product that hasn’t seen meaningful real-world usage yet, but with a fairly modest infra setup, it felt prudent. Given the latency (and variance) involved in AI agents processing a single request (multiple calls to LLMs, embedding models, etc.), it feels like any performance degradation is unlikely to be linear even with a small spike in concurrent requests. Makes me believe that load-testing will probably feature a lot earlier in the lifecycle of LLM-enabled products.