Step-by-step recipes for things I want to do the same way twice. Setting up a new Python research project, auditing a backtest for look-ahead, onboarding a new client site, scaffolding a Quartz wiki — the boring, repeatable work that benefits from being written down once.
The target reader is future me. If anyone else finds them useful, good.
Current playbooks
- Playwright E2E tests for a WordPress plugin — session caching, destructive-action defence at two layers, shortcode-map as the URL registry. The shape I run against WeCoza 3.0.
- A file-driven planning framework for AI-assisted coding — five phases, each phase writes a file, crash-resumable, mixed-model workflow (Opus → Gemini → Sonnet).
- Adversarial review via a second AI model — the structured prompt for piping Claude’s work to Gemini for a critical pass. Why different-model-family matters.
- Translating technical fixes for non-technical clients — two paragraphs, zero jargon, commit SHA, Trello tag. The template behind 14 months of WeCoza updates.
- Building a knowledge-graph research vault — iterative Obsidian-to-Quartz workflow with convergence detection. The shape behind SignalTrace and this CV.
Planned / likely next
- How I set up a new Python quant research project — stdlib-first, CLI entry points, snapshot hashing, honest postmortem on by default.
- How I audit an inherited backtest for look-ahead — checklist of the dozen places overfitting hides.
- How I onboard a new WordPress client — the first-week checklist that makes month-six tractable.
- How I scaffold a new Quartz wiki from scratch — without inheriting signaltrace’s CSS overrides (the mistake I made on this one).
Playbooks land here as they get written. Placeholder list above is where I keep track of what’s next.