All Articles

Read my recent thoughts on design, development, and the tech industry.
Two Herdr Mods: GitHub Tasks and Plan Usage
6 min read

Two Herdr Mods: GitHub Tasks and Plan Usage

Herdr lets you add native panels to its window with a little JavaScript. I added two, a Tasks panel that starts work from my GitHub project board and a strip showing my Claude and Codex plan usage, and here is how they work and what tripped me up.
My Mac Was Down to 32 GB Free. Here's Where 370 GB Was Hiding.
6 min read

My Mac Was Down to 32 GB Free. Here's Where 370 GB Was Hiding.

A developer Mac fills up with caches, build output, old tools and sound libraries you never open. What took the space on mine, the command that cleared each one, what I kept, and how I let an AI agent help without losing anything.
How I Tried Not to Fool Myself Backtesting Indian Markets
5 min read

How I Tried Not to Fool Myself Backtesting Indian Markets

My best trading strategy made ₹8.9 lakh on paper. A few honest checks turned it into a loss. Here are the checks, and the data problems I had to fix before any result meant anything.
I Tested ~150 Trading Strategies on Indian Markets. Only a Plain Index SIP Held Up.
5 min read

I Tested ~150 Trading Strategies on Indian Markets. Only a Plain Index SIP Held Up.

I tested day trading, swing trading, futures and options on five years of real NSE data, with every charge included. The strategies that looked great on old data stopped working on new data. A simple monthly SIP did as well as anything.
Can Indian OCR Read Music Notation?
6 min read

Can Indian OCR Read Music Notation?

I tested Bodhan's indic-ocr and Sarvam Vision against Claude Opus 5.5 on 27 pages of Bhatkhande notation. Both read Gurmukhi well and still lost by 35 points, because the music lives in marks that aren't letters.
What Happened When I Measured My Vision Pipeline
7 min read

What Happened When I Measured My Vision Pipeline

I had been saying "roughly 90% accurate" for months without measuring it. The eval showed that effort and model choice both move the result more than I expected, and that one config scored 59.4%, 59.4% and 90.6% on the same page.

© 2019-2026 Baljeet Singh. All rights reserved.