blog

How I pushed Qwen2.5-0.5B to 9,052 tok/s on a free Colab Tesla T4 in the H2Loop inference contest - from AWQ INT4 dead ends to FP16 plus model-free suffix speculative decoding.

How I built Logpipe - a small but real Vector + ClickHouse logging pipeline in Go. Why batching at an aggregator is the whole point, what 'Too many parts' actually means, and how disk buffers stop you from losing logs on a restart.

How I built TapTalk - a fully local, privacy-first speech-to-text app for macOS in Rust and SwiftUI. On-device Whisper, Core ML on the Neural Engine, chip-tuned inference, and a global push-to-talk hotkey. No cloud, no subscription.

How Swiggy, Airbnb, PhonePe, and Nubank ship UI changes without waiting for app store approval - and how I built a working BDUI + OTA system in React Native with Cloudflare Workers and EAS.

Dissecting the April 2026 Vercel breach - from infostealer malware to OAuth hijack to customer data exposure. Full kill chain, trust lessons, and survival guide.

How we built a production fair-value calculator on top of Dhan - Black-Scholes, cost of carry, live depth, and mispricing signals - for Saras and NSE F&O.