Skip to main content

Experiments

Small interactive tools I've built to answer specific questions, mostly about what AI models cost and how coding agents spend their context.

Agent Editor

Write or load a subagent definition, set every option with best-practice checks as you go, and export it for Claude Code or Codex.

Hook Explorer

See what every Claude Code and Codex hook event receives on stdin and can print on stdout, then check a payload of your own against it.

Skill Editor

Write or load a SKILL.md, set every frontmatter option with best-practice checks as you go, and export it for Claude Code or Codex.

Workflow Visualizer

Paste or load a Claude Code workflow script to see its agents, fan-outs, and phases as a diagram, check it for mistakes, and export it as a Codex workflow.

How Much Faster Is Fanning Out?

See how much faster a task finishes when you split it across subagents, and how much more it costs, as serial work and integration eat the speedup.

Is It the Model, or Your Machine?

Drop in your Claude Code session transcripts and find the failures that keep coming back, ranked by sessions affected, and how many come from your environment rather than the model.

Measurement Noise

Is the difference between two ways of working real, or just noise? The same five tasks, read two ways, show why it depends on how you measured.

Model Pricing Calculator

Compare what the same token usage costs across AI models, or drop in a Claude Code or Codex session to price its real usage on every model.

Where Your Context Window Goes

How much of a Claude Code context window is already spoken for before you’ve said anything, across five common setups, and which of those claims you can actually move.