Writing

You add an index and the query becomes instant. A week later, writes are slower, disk usage is up, and the planner ignores it anyway.

Paint mixing, rotating keys, prekeys, and the places encryption stops protecting you.

NVENC, x264, H.265, and why your GPU has a secret chip that encodes video without touching a single CUDA core.

Segments, playlists, and adaptive bitrate. How one video becomes hundreds of files before you press play.

Learn how the Origin Private File System stores large browser files, supports random access, and enables SQLite and worker-based storage.

You tap a button and money teleports. Here's what actually happens in those 3 seconds.

A tiny container runtime built in Go to understand namespaces, chroot, proc mounts, and basic cgroup v2 limits.

You type a few letters, suggestions appear. Behind that is a trie, then ranking, then a mess of heuristics.

Why reliable undo and redo require command inversion, grouping, state dependencies, branching history, and special handling for collaboration.

There are many valid diffs for the same change. Myers just picks the first one it finds.

Real-time collaborative editing looks impossible until you stop syncing state and start syncing intent.

I thought diff was a simple loop. It's actually a shortest-path problem.

Learn how JPEG and PNG compression turn raw pixels into smaller files using transforms, prediction, LZ77, and Huffman coding.

Why I built this personal software blog and what I plan to write about: engineering, products, experiments, tools, and practical lessons.

A minimal implementation of React Query from scratch. Caching, tag-based invalidation, and automatic revalidation in ~200 lines.

A zero-dependency client-side router built with React Context and the History API. ~100 lines of code.

A minimal global state management library inspired by Zustand, built to understand how state management works under the hood.

A ~100 line virtual DOM implementation. Virtual nodes, rendering, and diffing from scratch in plain JavaScript.

An interactive tool for understanding how language models break text into tokens. Visualize tokenization, inspect token metadata, and learn why it matters.