One word can wreck a whole explanation.

The exact clicks

1. Read the line out loud, exactly as written.
2. Ask someone else — or yourself, out loud, in a different sitting — "what did I just say, in your own words?"
3. Never accept a yes-or-no answer to "did that make sense?" — it doesn't catch anything; a paraphrase does.
4. If the answer doesn't come back matching what the line actually meant, the line isn't done — rewrite it, don't just slow down on it.
5. Repeat for every line that names a tool, a technical term, or an interface element before it gets recorded or published.
6. To run this automatically in Claude Code: open /s/explain-steps on this site, download SKILL.md, and put it at .claude/skills/explain-steps/SKILL.md in your own project — Claude Code picks it up with no restart and no config.

What changes

• The real example this rule caught: "Add this to the column" (said once, never explained) became "Instagram fills this box on its own" — same fact, zero required vocabulary.
• The rule lives in a written, free skill called explain-steps — the same one behind every step-by-step explanation this channel makes, not invented just for this episode.
• It's a five-minute check before recording, not a rewrite of the whole production process — the cost is asking the question, not redoing the work.

What it will not do

• This catches whether a sentence is understandable, not whether it's factually correct — a wrong but clearly-worded claim still needs separate fact-checking (see episode 15's own correction).
• It works best with a second person; testing it on yourself only works if you can genuinely forget what you meant to say, which is harder than it sounds.
• This rule caught real gaps in the back catalog too (episodes 3, 6, 7, 9, 10, 11 all had unexplained jargon or steps too vague to actually follow) — those have since been rewritten to the same standard, not left as an exception.