23 strategies · 5 tiers · every claim sourced
You pay for context size × turns, every turn, forever. Most "save tokens" advice is vibes. This is the opposite: 23 strategies, each with the cost mechanic behind it, the exact command to copy, an honest impact rating, and when NOT to use it. One interactive HTML file — open it, search it, copy from it.
Instant download · no install · works offline · copy button on every command
The problem
"Just be concise" doesn't tell you which lever moves the bill, by how much, or when it costs you quality. You need the mechanic, not the platitude.
What it does
Start at Tier 1 — most people never need to go past it. Each tier goes deeper for when you want more.
One-task-per-session, /rewind over /compact, pick your model at session start, targeted reads, and measuring with /context.
Trim your CLAUDE.md, use permissions.deny read-rules, cut MCP servers you don't use, and prune code-intel plugins.
/effort, thinking budgets, background-traffic toggles, and auto-compaction tuning — the settings most people never touch.
Convert PDF→Markdown before sending, resize screenshots, and filter giant tool output with a hook before it ever hits context.
Fork subagents that share the cache, route cheap work to smaller models, use plan mode, and run async jobs on the Batch API at 50% off.
What's in the box
Start at Tier 1 today
23 sourced strategies, one searchable file, a copy button on everything. Most people get the biggest win from Tier 1 alone.
Instant download · 30-day Gumroad refund policy applies
Setting up Claude Code from scratch? Start with the Claude Code Setup Playbook — $29.
See all my Claude Code tools → expressive446.gumroad.com