I tracked what works and what doesn't when trying to reduce Opus 4.6 spend. Model routing, session practices, the advisor tool, compaction.
Blog
-
What actually cuts Opus token costs
-
Per-tab proxy routing in Chrome
A Chrome extension that routes individual tabs through a SOCKS5 proxy using PAC scripts and an optional control API.
-
Surfacing Claude Code's rate limit data
Claude Code keeps your API rate limits in memory but only shows a warning near 100%. A script to pipe those limits to your status line.
-
My home server setup
How a transcription script turned into 11 self-hosted services.
-
I needed a rental, so I built a listing aggregator
Building a real estate aggregator for Santa Fe with a browser extension, Qwen vision API, and a static map.
-
Making Claude Code background tasks observable
A hook-driven pipeline to capture background task output and inject it back into the assistant's context.
-
Persistent memory for Claude Code with codemem
A local memory engine for AI coding assistants. Setup, daily use, and some notes from contributing.
-
Better Docker logs with Python
A Python script for reading Docker logs with colors, search and scroll.