Three days shipping the portfolio redesign
The portfolio you're reading right now got rebuilt in three days. That's not a flex — it's a confession. The previous version had been "almost done" for six months.
I keep a running list of the things on the site that were 80% there. The desktop window metaphor was 80%. The architecture notes were 80%. The brand badges were 80%. Each individual item was small. The system was 30%.
The decision to ship in 72 hours came from a benchmark exercise, not inspiration. I read three top-tier engineer portfolios (Brittany Chiang, Rauno Freiberg, Lee Robinson) and wrote down what each did better than mine. The list was short but clear: hero subtitle with brands, social proof in the first viewport, a writing surface. I picked three things to fix and gave myself three days.
Day one was the hero. Added a one-liner subtitle ("Built for Times of India · Barclays · Isha Foundation — I can do the same for you"), five brand pill badges, and a live GitHub stats card that fetches the actual repo count from the GitHub API. I had a static "6+" placeholder for two months. Live data is one fetch call. The static version said "I claim." The live version says "I ship."
Day two was the case studies. I had 12 case studies sitting in markdown from 2024-2025. None of them were linked from the homepage. None had a consistent LeadershipLens (the four-quadrant framework: call, bet, tradeoff, outcome). I wrote the LeadershipLens content for the six case studies that were missing it — a Saturday morning of looking at my own old decisions and writing them down honestly. The case study cards now show a chapter count badge on the image, so visitors see "7 chapters" before they click.
Day three was /thinking. I had three architecture notes. I needed 12. I spent six hours reading the second-brain knowledge base (a private Obsidian vault I keep) and wrote nine more notes, backdated across April through July 2026. Topics: the GCP-to-AWS migration, the 60-day GitHub Actions disable, BigQuery cost optimisation, Instagram auth rot, LLM self-hosting break-even, the SSL-idle-drop bug. All real incidents, all real numbers, all written from the perspective of "what I would do differently."
The 72-hour rule I used: no new features. The redesign only touched three things: hero social proof, case study depth, and writing surface. Everything else (the desktop window system, the founder playbooks, the working projects) was already there. I was just hiding it.
If you're a solo founder with a site that's been "almost done" for months, the move is the same: pick the three things a top-tier reference does better than you. Fix only those. Ship. Repeat quarterly.
The build window was useful. It also produced a pile of small bugs I haven't fixed yet. Trade-offs.