Thoughts on engineering, design, and building products.
How a simple LeetCode Easy interview question became a distributed systems crash course and evolved into a distributed systems nightmare involving 100TB of data, Spark, and Bloom Filters.
Standard shell history is useless for debugging because it lacks context. I hacked Zsh to track execution time, git branches, and exit codes for every command—creating a permanent "flight recorder" for my engineering work.
Running local LLMs in WSL eats up 50% of your RAM. The fix? Run Ollama on Windows and bridge the connection. Here's the one-line script that solved my dynamic IP headaches.
Running three production services on Vercel's free tier, I noticed response times ranging from 800ms to 25 seconds. Through data-driven optimization and strategic use of cron-job.org, I achieved consistent sub-second responses across all services—without spending a dime. Here's how I measured, analyzed, and optimized cold starts for Node.js and Python serverless functions.
When my portfolio went down on New Year's Eve due to free-tier limits, I didn't just pay up, I engineered a multi-cloud failover system using Vercel and Render that costs $0/month. Here's the story of how I migrated my Python backend to Serverless and built a client-side circuit breaker.
Google Antigravity not working on WSL? This guide shows you how to fix Linux script execution, enable the browser subagent, and automate the entire setup. No more "invalid command" errors!
