x-content-playbook

X Content Playbook — performance-grounded rules

Source: full-population teardown of the founder's account, 1 Jul – 31 Aug 2026. 11 original posts, ranked by views / comments (replies) / saves (bookmarks). Median reach ~465K; outlier floor 914. This is measured data, not theory — when a rule below conflicts with intuition, the rule wins until re-measured.

Refresh rule: this snapshot ages. Re-run the same teardown (X date-filtered search, all originals, views/comments/saves) at the start of each quarter and update the numbers and rules here. Rules derived from Jul–Aug 2026 hold until then.

---

RULE 0 — pick the target metric BEFORE writing

The single most reusable finding: the three metrics have three different levers, and they are nearly independent.

You want	The lever	Evidence
Reach (views)	news peg + category-killer line, or a top-4 hook	top 3 posts all ride a launch or a labour reframe (1.88M / 1.35M / 1.32M)
Replies (leads)	a "Comment [WORD]" gate — replies are engineered, not earned	remove the gate and replies collapse to single/low-double digits even at 1M+ views (Aug 20/24/26)
Saves (intent)	teach a repeatable, portable move — a named concept or playbook	saves cluster on "white space" (147), the 48x playbook (93), the launch (1,581)

A post that doesn't know which of the three it's chasing will get none of them. Decide first; the shape follows.

RULE 1 — the hook does ~90% of the work

Every winner opens with one of four hook families. The body rarely changes the outcome; the first line does.

1. Bold reframe / category kill — "Your whole ad team is now a single hire." · "The report era of AI just ended."
2. Metaphor / curiosity gap — "Your best salesperson sells from a room nobody visits."
3. In-scene story or dialogue — open mid-scene on a Slack ping ("…Found a 4th nobody's touched. Build it?") or a founder moment ("woke up to notifications stacked on her lock screen").
4. Contrarian reversal — "max talked them out of showing skin at all." · "their own product never appears in it." Strongest family for cheap reach in the dataset — lead every teardown with it.

A flat factual announcement is quiet. The announcement welded to a one-line worldview is what travels.

RULE 2 — the copy skeleton (repeatable)

1. Line 1: the hook (one of the four families) — carries the scroll-stop.
2. White space, then a 1–2 line problem or contrast, with a number — 347 ads, $18 CPA, 3.2×, 48x, 11 bottles. A concrete figure in the first three lines buys credibility before the claim. (Numbers must be real or clearly illustrative — never invented performance figures; run-times, counts and structural numbers are the safe kind.)
3. A scannable → bullet list — what the agent did or understood. Never paragraphs.
4. A provocative reframe closer — often a signature refrain.
5. CTA only if the goal is replies/leads (Rule 0). Omit on pure-reach teardowns — but see Rule 5.

RULE 3 — the product is a character

"Max" + the pronoun "he" recur across the best posts. The persona is the account's single most differentiated asset: hireable-teammate framing, Slack pings, day-in-the-life scenes. Personified SaaS is a story; features are not. (Consistent with the locked hire-voice: he/him after the reveal, hire verbs, never "it generated".)

RULE 4 — signature refrains compound

"Marketers used to be people. Your next one won't be." reliably earns reposts and saves across multiple posts. Build a bank of refrains so reposts have more than one recognisable line to latch onto. A refrain is a portable closer: short, quotable, slightly uncomfortable, and repeated verbatim across posts.

Current bank:

• "Marketers used to be people. Your next one won't be."
• "You sleep. Max ships."
• Candidates should be added here as they prove out (a refrain "proves out" when it appears in quote-tweets without prompting).

RULE 5 — the CTA decision table

• Launch / persona / integration posts: comment-gate + a real offer ("Comment MAX — first 50 brands get…"). Drove 193 and 68 replies.
• "Hidden money in your account" posts: always gate — personal + gate is the highest reply-rate combo (Aug 25: 66 replies at 318K views).
• Teardown teasers: historically ungated and it cost us — Aug 20/24/26 pulled 200K–1M views with single-digit replies. Add "Comment TEARDOWN and I'll send the full breakdown" — converts lurkers at no cost to the hook.
• Every gate word must have its payload built before the post runs, and each post gets a distinct word so replies are attributable.

RULE 6 — the angle families in rotation

Rotate; never run one family two posts straight.

Family	Posts	What it buys
Launch & integration (news-pegged)	Aug 5, Jul 22, Jul 30, Jul 9	biggest reach + saves
Customer creative teardown (curiosity teaser → thread)	Aug 28, 26, 24, 20	high views, low replies unless gated
Hidden money in your account (personal + gate)	Aug 25, Jul 1	replies/leads

RULE 7 — the outlier rule (what never runs standalone)

Jul 11 ("the most mission-critical feature… is Ad previews") collapsed to 914 views vs a ~465K median. It was a founder-philosophy essay: no customer outcome, no visual, no persona, no "you" benefit up top, no CTA.

The lesson: on this account, craft does not buy reach. Customer outcomes, bold reframes and a visual do. Founder-philosophy content is allowed only:

• reformatted behind an outcome hook + a picture, or
• run as a reply under a performing post, never standalone.

RULE 8 — news pegs are the cheapest reach

Ship a launch-style post within 24h of every major AI-platform moment (Claude and ChatGPT news were the two largest reach events of the period). Borrowed relevance beats owned cleverness. This pairs with the news-reactive article lane — the X post is the fast twin of the Friday article.

RULE 9 — cadence

One post every ~5 days worked. The two biggest gaps (mid-July, mid-August) sat right before the account's best months — more at-bats at this hit rate is pure upside. When planning a calendar, treat a >6-day gap as a defect.

---

Pre-publish checklist (run on every draft)

• Target metric chosen (reach / replies / saves) and the shape matches it
• Hook is one of the four families; the first line survives without the rest
• A concrete number in the first three lines (real or clearly illustrative)
• → bullets, no paragraphs; white-space-heavy
• Max is "he", framed as a teammate, if the product appears
• Closer is a refrain or a reframe, not a summary
• CTA: gate word present iff goal = replies; word is unique; payload exists
• Not a standalone founder-philosophy essay (Rule 7)
• No invented performance numbers (permission + 2.5–4× range for real ones)
• ≤5 days since the last post, or this one closes the gap

Calendar-planning rules (when planning a week/fortnight)

• Assign each slot a target metric first (Rule 0), then pick the concept.
• Rotate the three angle families (Rule 6); never two of one family in a row.
• Reserve capacity for one unplanned news-peg post per week (Rule 8).
• Teardowns get gates now (Rule 5) — plan the payload as part of the slot.
• Grounded concepts beat invented ones in every selection round to date — source from real sessions (Watch Max Work), real ads, real data.

Boundaries with sibling skills

• launch-post owns launch-post structure and hook iteration protocol — this skill supplies the performance rules it must satisfy.
• format-radar owns teardown episode production — this skill decides the teardown's CTA and metric target.
• The judgment-over-volume rule (house rule: judgment over volume) still governs WHAT concepts run; this skill governs HOW they're written and scheduled.