People tend to read less in general, and much less in correlating or analysing. The recent discussion around the cautrāmnāyas has triggered lay smārtas a bit too much. The veracity remains to be discussed and their own internal discord on timelines for śaṁkara still not resolved. Amidst all this, I do see a repeated quotation. ś i v ā y a v i ṣ ṇ u r ū p ā y a . . . Where does this come from anyway? ...

As part of making small changes each day, chose to pick up a small implementation with Redis streams today. The idea was to pick up something I have not explored rather than sharpen the same toolkit. At the outset, the question was why Redis streams and why not use PubSub itself. This question came up because of the frequency with which I use PubSub and event driven architectures in general. A common case would be say, an order service that talks to a payments service, and then to a procurement service for interfacing with vendors. ...

Does 70MB sound like a lot? “It depends” is a great answer. On my VPS that powers crisp, where I have 20 process runners and three web/ping servers, that’s a steal. Even when the processes themselves are lean enough to be both Tamirat Tola and Usain Bolt, that’s a welcome win. But that was somehow not the thought when I embarked on a change late this evening, but ended up a sweet surprise as I eased into the night. ...

Late last weekend was a minor pivot. One of the core limbs of the erstwhile crisp app was a content extractor written in Javascript. A certain something that wasn’t exactly fast to my liking then, and much more so now when I am chasing milliseconds for entire operations. So in the new iteration, broke it down to its finest bits and started moulding from the ground up. In every sense, this was a rebuild from scratch with zero parallels to the historical one — in philosophy, pace, or being a panacea of sorts. ...

I run 19 instances of a C program that fetches RSS feeds on a 1vCPU VPS. Each run is short, but over time, the CPU profile felt off – spiky, higher than it should’ve been. Turns out, the fetch loop was subtly blocking. No threads, just enough delay to cause churn. I rewrote it to run on an event loop – non-blocking I/O, no busy waits. After the change, CPU usage dropped by half. No infra changes, no concurrency hacks. Just better rhythm. ...

For nearly two decades in tech, I’ve quietly built things that mattered – writing bash scripts late after class, and later, quietly rebuilding Crisp’s infrastructure almost from scratch and at near-zero cost. But as a CTO and solo builder, I’ve realized that quiet craftsmanship alone keeps your journey and your impact invisible. Invisible work doesn’t build community, influence, or the momentum needed to lead and inspire. After years of encouraging founders to “share their process,” I knew it was time to walk the talk. ...

A quick nod to my October 2024 post where I wrote about a resurrection; a spark of revival just a week before I joined my current role. After that, almost a year of quiet, no major moves, just the hum of the background. But now, here I am, with something serious finally done. This post is the story of that journey: from stillness to momentum, from a borrowed MacBook Air to a fleet of blazing-fast containers. It’s about persistence, evolution, and the indie CTO spirit that never quits. ...

[tl; dr] I spent some time in the recent past weeks to pull crisp back up, to the extent possible and incrementally adding back more from its older avatar. All that’s below is a brain dump of whatever has encountered over the phase. Something a long time due has been crisp's resurrection; or what it looks to be today. The wonderful thing about a boring job, a boring persona, and any such is the element of surprise that it throws to those that never saw things coming. The attempted super-villain-esque jot as a joke aside, it has always been a focus to build things that people don’t generally associate me with. Be it picking the mic up for a sort of impromptu singing at work, or over a few nights rebuilding some parts of crisp from memory, this has been snarky to the insider. ...

This evening has been a good one. A welcome break if I may, from a few weeks without enough to feed the brain. Of course, a minor detour existed with refreshing logarithms and some trigonometry. Cutting back, stumbled upon some excellent content on Distributed Hash Tables almost by accident. The Kademlia DHT is just about old enough for its first job post a bachelors. The genius of the trick lies in the binary of XOR for distances and a sort of binary tree space for retaining information about neighbouring nodes. Considering Kademlia focuses on distributed hash tables, one of the first concerns is about calling back and identifying nodes prior to fetching data. ...