Cursor AI Spend by Repo and Model
How to see which repositories and models drive your Cursor bill — local cost tracking, repo attribution, and comparison to Claude Code on the same machine.
blog
Field notes on AI-assisted development.
How Claude Code actually works under the hood — transcripts, skills, customization layers — and what to do with that knowledge.
-
-
Track Claude Code Token Usage and Cost (Beyond /usage and ccusage)
Compare PromptConduit, Claude Code /usage, ccusage, and the Anthropic dashboard for tracking token spend — per turn, per model, with cache breakdown and repo context.
-
How to Write a Claude Code Skill From Scratch (With a Real Example)
Step-by-step guide to writing your first Claude Code skill: the SKILL.md format, frontmatter fields, when to use project vs user scope, and a complete /review skill you can copy.
-
Claude Code Has 5 Ways to Customize It — Here's When to Use Each
A complete guide to Claude Code customization: CLAUDE.md, Skills, slash commands, hooks, and memory. Where each one lives, when to use it, and how they overlap.
-
Where Claude Code Stores Your Transcripts (and How to Actually Use Them)
Claude Code writes every session to a JSONL file on disk. Here's exactly where, what's in each file, and what you can do with that data — grep, replay, back up, or turn it into reusable skills.