Writing

A selection from 64+ published articles, hosted here. Most also run on Medium as syndicated copies.

  • The AI Demo Worked. Now Show Me the Cleanup.

    2026-10-26

    One prompt, more code and a mountain of tokens tell you surprisingly little about what your team can ship.

  • Stop Teaching Your AI Agent the Wrong Lessons

    2026-10-21

    How to decide what an agent should remember, where it belongs, and whether it helps the next task.

  • See Where Your AI Agent’s Answer Came From

    2026-10-19

    A hands-on Atomic Agents and Phoenix guide: run a planner, calculator and answer writer, then inspect the real trace.

  • Stop Making Claude Code Rediscover Your Codebase

    2026-10-14

    Give each new session a useful starting point, with a compact project index, source pointers and a runnable example.

  • The Plugin That Turns Your Claude Code into a Continuously Self-Improving Monster

    2026-09-11

    Quartermaster sets up Claude Code for a new or existing project, then watches real sessions, proposes one approved improvement at a time, and checks whether its last recommendation actually helped. It installs and wires codebase-mapper, live-rules, stack plugins, permissions, and whatever else the workspace needs.

  • Why Using Astra Inside Claude Code Is the New Meta (& How To Do It)

    2026-09-07

    OpenAI shipped GPT-6 Astra on Thursday. By Friday it was the model driving my Claude Code sessions, on the ChatGPT subscription I already pay for, with nothing else in my setup moved. The frontier model and the harness are two separate decisions now, and this is how the split works day to day.

  • Meet Sidequest, the Kanban Board Where the Cards Work Themselves

    2026-09-02

    One expensive model plans while cheaper ones do the work in parallel, and everyone is duct-taping their own version out of prompts and shell scripts. Sidequest is my version: a free Claude Code plugin where side issues become tickets, each kind of work picks its own model, and executors run in their own worktrees with a pinned verifier.

  • How to Make Claude Code Do Things Its Own Docs Don't Mention

    2026-08-19

    A plan of mine hinged on "a startup hook can only add context". The installed Claude Code binary said otherwise, and the docs would not say so for weeks. Docs lag, omit, and describe the version the writer had; the minified bundle on your disk is an exact description of what you are running, and Claude reads it like prose. How I grep it, store the finds by version, and where the method stops.

  • How a Quick Sidequest Turned into the Most Powerful Claude Code Plugin I Have Ever Used

    2026-08-18

    It started as a list for the "oh, by the way" thoughts that show up mid-task, so the main session could keep going. It grew, one obvious step at a time, into a local ticket board where each kind of work carries the model that fits it, workers run in their own copy of the repo, and my ChatGPT subscription is on the roster next to Claude.

  • The Quest for Local Model Routing Inside Claude Code Starts with a Codex Gateway Plugin

    2026-08-07

    Claude Code could only reach Anthropic models, so I was picking from a third of my real options. model-gateway is a plugin that puts the GPT and Grok models from my existing subscriptions straight into the /model picker, keeps itself updated, and leaves Claude traffic going to Anthropic on my own credential.

Want this kind of analysis in your inbox once a month?

Your email goes to Buttondown. Unsubscribe any time. Privacy notice.