Question 1

What's your take on the degree vs. skills debate?

Accepted Answer

I'm not a 'degrees don't matter' evangelist — for many people the structure and the credential are genuinely worth it. My own bet was to let demonstrated work carry the signal: by the final year I'd already shipped a deep body of production work, and the highest-leverage use of my time was building more of it. The practical reality is that degrees still filter resumes at companies that haven't found better signal — that's a real cost I accept. My answer is to make the work undeniable: 15+ production projects, open-source tooling, and this site. Some companies won't interview me regardless, and that's fine — the ones I want to work at care more about a production system I built from scratch than the line on the form.

Answered about 1 month ago

Question 2

How do you manage working solo on projects that teams of 10 would normally staff?

Accepted Answer

Aggressively narrow scope, ruthlessly reuse infrastructure, and batch context switches. For scope: I ship the smallest version that proves the concept, then expand. For infrastructure: that shared Docker mesh exists specifically so I don't reconfigure Postgres every project. For context: I block 4-hour deep work sessions and avoid switching projects mid-session — the mental overhead of re-loading context is real and underestimated. The honest answer is I also just work a lot. The OTT platform was 8 platforms because the client needed 8 platforms, not because I enjoy building tvOS apps.

Question 3

213 WPM — is that real? What keyboard layout?

Accepted Answer

Real, verified on MonkeyType. I designed a custom layout called X-Type in 2015 — but not by tweaking an existing layout. I built a custom corpus from my actual writing (Hindi, Hinglish, and English), uploaded it to the Patorjk Keyboard Layout Analyzer, and designed the arrangement from scratch against that corpus's real character and bigram frequencies. On my own text it beat QWERTY, Dvorak, Colemak, and Workman on every metric the analyzer measured. The core insight is that most typing speed gains past ~120 WPM come from reducing same-finger bigrams and maximizing home row usage — not just raw practice. The awkward part: I can't type on a standard QWERTY keyboard above ~90 WPM anymore, so every pair-programming session requires me to switch layout on the other person's machine.

Answered about 2 months ago

Question 4

Why Rust? You already know TypeScript deeply — what made you invest in it?

Accepted Answer

Two things pushed me. First, I was building Docsee (a Docker GUI) and Go felt too opinionated, Python too slow for a native desktop app. Rust gave me the Tauri ecosystem and sub-50ms cold starts without an Electron runtime. Second: I kept running into problems where I wanted to reason about memory and performance at a level TypeScript simply can't express. The borrow checker is brutal for the first few months but after that it becomes genuinely addictive — the compiler catches whole categories of bugs at compile time that I used to find at 2am in production. Now I write Rust for anything perf-sensitive or long-running: CLI tools, TUI apps, parsers.

Answered about 2 months ago

Question 5

What's the hardest architectural decision you've had to make in production?

Accepted Answer

Moving the Prachyam mail infrastructure off a managed SMTP provider to self-hosted Mailcow mid-flight, with active users. We were at ~200K emails/month and the provider costs were becoming unsustainable. The hard part wasn't the technical migration — it was the IP reputation bootstrapping period. You can't just spin up a new mail server and expect ~20M emails over a campaign to deliver. I spent two weeks on DNS configuration (SPF, DKIM, DMARC across 12 domains), another week on gradual warm-up, and had to build a priority queue system after hitting a race condition that caused duplicate sends. Outcome: deliverability improved and costs dropped significantly. But that two-week risk window is not something I'd do again without more staging runway.

Answered 2 months ago

Question 6

How are you running 116 containers on a single MacBook without it melting?

Accepted Answer

The 116 number is the total count across 19 Docker Compose profiles — on any given working session only 30-40 are actually running. The key insight was treating services as shared infrastructure instead of per-project: one Postgres, one Redis-compatible (Dragonfly), one MinIO, shared across every project. I also use named profiles aggressively — docker compose --profile core --profile ai up -d, and everything else stays down. The M1 Max with 64GB is genuinely overspecced for this. The thermal situation is fine unless I'm running Ollama + full monitoring stack + video transcoding simultaneously.

Answered 2 months ago

Question 7

You started coding at 9. What was the actual first thing you built?

Accepted Answer

A 3D car racing game in 3D RAD — a node-based game engine. No tutorial. I figured out collision detection by breaking it repeatedly, built custom terrain meshes and a skybox, and wired up enemy AI with three difficulty levels and a track editor. It actually worked. The feeling of making something move on screen at 9 years old is something I still chase.

Answered 3 months ago

Ask Me Anything

Pending

Answered7

Ask a question