Writing a 994k-subscriber newsletter with Claude Code: cutting 8 of 10 words to train 74 rules
rubenhassid · x · 2026-10-27
rubenhassid, author of the AI newsletter "How to AI" (994,381 subscribers), published his full writing pipeline built on Claude Code + GitHub. His latest issue was drafted by Claude Opus at 2,303 words; he kept 438 verbatim — cutting roughly 8 of every 10 words.
Key mechanics from his flow diagram:
- rules.md holds 74 rules, each derived from a real edit he made; rules that never fire get deleted
- Writer/Checker split: a writer Claude sees only the outline and 2–3 past issues; a fresh checker Claude reads rules.md, quotes every violated line, and never rewrites
- Blind test each issue: his version vs Claude's draft; if the checker can't pick his, the rules get fixed
- edits-log.md tracks 81 corrections in 14 days, CLAUDE.md is compressed to 207 words, and diffpublished.py compares sent vs drafted text
He says it's reproducible with zero coding experience — Claude will tell you what it needs.
More from coding & agent
- 'Codex decided': devs admit making up reasons in code review for agent-written code — FlatStill2540 · 2026-10-27
- Running AI agents on a legacy IBM WMS: enforce tool order, script checks, verify evidence — ksrijith · 2026-10-27
- 365 public MCP servers tested: 51% lack tool input descriptions, the most common bug — abhiramreddykalluru · 2026-10-27
- CoreWeave ARIA agent ships four new live in-chat panels for ML run analysis — _ScottCondron · 2026-10-27
- opentel-mcp brings OpenTelemetry silent-failure detection to Python MCP servers — Thirumalaiboobathi · 2026-10-27
- A hands-on local-model benchmark: porting an old screensaver to Python OpenGL with Qwen 3.8 — prusswan · 2026-10-27