/blog

AI & agents

Building with Claude Code and other agents, and owning the work they help make.

  1. AI & agents 11 min read

    Rebuilding This Site in One Claude Code Session

    I rebuilt codywilliamson.com in one 15-hour Claude Code session: about 5,000 tool calls, 30 merged PRs, and 33 messages from me. What those messages did, and what went wrong on both sides.

  2. AI & agents 7 min read

    Porthole: A Phone Window Into My Claude Code Sessions

    I run multiple Claude Code sessions across tmux windows and wanted to read them — and reply to them — from my phone. So I had Claude Fable 5 build the tool, orchestrating a fleet of Sonnet and Opus sub-agents, while friends watched the commits land in real time.

  3. AI & agents 14 min read

    Revisiting Autonomous Development with Opus 4.8

    I handed Claude Opus 4.8 a build spec and got back a working, tested, published tool in one session. Here's the framework that made a long autonomous run reliable instead of reckless.

  4. AI & agents 14 min read

    Autonomous VPS Monitoring with Claude Agent SDK

    An AI agent that runs twice daily, performs VPS health checks, applies updates, improves its own tooling, and reports to Discord.

  5. AI & agents 3 min read

    I Didn't Want Copilot to Be the Odd One Out

    How I adapted an existing Codex plugin pattern to bring GitHub Copilot into Claude Code as a first-class citizen — no context switching required.

  6. AI & agents 7 min read

    AI and the Problem of Responsibility Laundering

    AI is a powerful tool that transforms how we work. But when engineers start citing ChatGPT as their authority, we have a problem. This is about intellectual ownership and why abdication erodes trust.