FREE SETUP KIT · VERIFIED AUG 16

The Unlimited Setup — Build All Day on One Claude Plan: the Paste-In Token Discipline Block, the 12-Fix Checklist Ranked, and the Window Playbook

The companion to the video. Most of your Claude allowance isn't spent on answers — it's spent on re-reading, re-explaining, and restating, and this kit is the 12-fix ladder that stops the leak. PAGE 2 — THE PASTE-IN: a copy-ready Token Discipline block for your CLAUDE.md — one paste that tells Claude to stop narrating, stop echoing files back, and stop re-reading what it already read; output tokens cost 5× input on every current tier, so this one block attacks the most expensive text in your session. PAGE 3 — THE 12-FIX CHECKLIST, ranked 12→1 from the two-minute habit to the biggest lever: read the meter with /usage and /context, /clear between tasks to kill the all-day session, scope your asks, reference paths instead of pasting files, compact warm at natural breaks, keep CLAUDE.md under 200 lines with workflows moved into skills, pick model + effort at the top of the session so you never bust the cache mid-task, subagents as context firewalls, model: haiku for simple subagents, workhorse model by default, batch heavy work inside one 5-hour window — and the paste-in as fix #1. PAGE 4 — THE WINDOW PLAYBOOK: how the limits actually meter you — the rolling 5-hour window, the weekly cap across all models, the separate big-model weekly cap with its /model escape hatch, the one shared pool across web, mobile, and Claude Code, why model-switching doesn't reset anything, why no exact token numbers are published, and the usage-credits escape valve for deadlines that can't wait. PAGE 5 — OUR RECEIPTS, measured on our own machine August 16, 2026: 36,116 input-side tokens moved before a one-word answer, 37.7k tokens used at session start before any work, and the MCP on/off delta of 45 tokens that retires the old 'turn off your MCPs' advice. Every official claim verified against Anthropic's support and docs pages the same day. Independent kit from Hyperautomation Labs — not affiliated with Anthropic.

Subscribe to Hyperautomation AI ReportGet the PDF freeKeyword: UNLIMITED

Free. No spam. Unsubscribe anytime.