A Friend — Distribution

references__aesthetic-register-anti-slick

9 min read

Synced from references/aesthetic-register-anti-slick.md in content-extraction skill on 2026-05-18. Edit upstream in the skill; this file is overwritten on next sync.

Aesthetic Register — Anti-Slick Guard

The brand register guard the virality sub-agent applies. A Friend’s voice is punchy, sharp, real, essence-loaded, considered, perceptive, illuminated — explicitly NOT motivational-speaker, NOT hustle-bro, NOT productivity-influencer, NOT TED-pacing, NOT Gary-Vee energy, NOT brain-rot bait. This file makes the seven adjectives Daniel named (verbatim 2026-05-03) operational, and names the failure modes the sub-agent must detect.

The premise: short-form virality data tells you what algorithms reward. Brand register tells you which of those tactics A Friend can use without breaking voice. The sub-agent has BOTH knowledge — but when they conflict, brand wins.

The seven-axis register (Daniel verbatim, 2026-05-03)

“Punchy, sharp, real, essence-loaded, considered, perceptive, illuminated.”

These are the seven axes A Friend clips are scored against. Each is independent — a clip can score well on some, poorly on others. Two or more weak axes flag the clip for register-drift review.

1. Punchy

What it is: the line lands in 1-3 seconds. Stand-alone declarative. Lean syntax. Earned weight.

What it’s NOT:

  • Long setup before payoff
  • Buzzword stacking (“revolutionary AI-powered companion”)
  • “What I’m trying to say is…” preamble
  • Filler-fronted sentences

Score:

  • ✓ — every clip-second-0 sentence works as a stand-alone tweet
  • ⚠ — opener needs 4-5s to land (tolerable but flag)
  • ✗ — opener fails to be self-contained; viewer needs context that doesn’t arrive

2. Sharp

What it is: specificity over abstraction. Named decisions, named tools, named numbers. No vague gestures.

What it’s NOT:

  • “Various tools” / “lots of agents” / “a bunch of things”
  • “Recently” / “the other day” / “sometimes” — temporal vagueness
  • “Many people” / “a lot of founders” — agentless attribution
  • “Pretty” / “really” / “actually” / “basically” — softening modifiers

Score:

  • ✓ — every named noun is concrete; every modifier is earned
  • ⚠ — 1-2 vague terms in surface but specifics elsewhere
  • ✗ — clip leans on abstraction; nothing concrete pulls weight

3. Real

What it is: earned vulnerability. Specific failure / specific stake / specific moment. Not performance.

What it’s NOT:

  • “I cried last night” — performed vulnerability
  • “Building a startup is hard” — generic struggle
  • “Let me be honest with you” — meta-performative honesty
  • “We’ve all been there” — false-collective

Score:

  • ✓ — vulnerability is named, dated, specific (segment references “today I woke up”, “three months in”, “for the third time”)
  • ⚠ — vulnerability gestures at specific without naming it
  • ✗ — humble-brag pattern OR generic struggle without stake

4. Essence-loaded

What it is: every kept second carries weight. Substance density per second is high. The kernel (I-believe-that) lives somewhere in the clip and the surrounding material supports it.

What it’s NOT:

  • Filler that doesn’t advance the argument
  • Repetition without rhetorical purpose
  • Topic drift mid-clip
  • Multiple competing arguments in one clip

Score:

  • ✓ — every kept segment supports the clip’s Essenz; no filler
  • ⚠ — 1-2 segments are softer but the spine is strong
  • ✗ — clip argues multiple things OR has visible filler

5. Considered

What it is: the clip feels REHEARSED-NATURALLY. Not improvised, not scripted. Daniel thought about this before, but he’s not reading. Cadence is intentional, choices are deliberate.

What it’s NOT:

  • Improvised rambling that hits good lines by accident
  • Reading-from-script energy (over-articulation, paragraph-pacing)
  • Stream-of-consciousness without arc
  • Frantic / amped delivery

Score:

  • ✓ — sentence structure varies, cadence has rhythm, beats land where they should
  • ⚠ — some segments feel improv (acceptable if rest is considered)
  • ✗ — entire clip reads as either improv-only OR script-only

Sub-agent heuristic: sentence-length variance + presence of two-beat structure + named-decision count. High variance + ≥1 two-beat + ≥2 named decisions ≈ considered.

6. Perceptive

What it is: Daniel SEES something most people miss. Insight that hits because it reframes existing assumptions. Not received wisdom.

What it’s NOT:

  • Productivity advice (“waking up early changed my life”)
  • Generic founder-stake (“the journey is the reward”)
  • Re-said wisdom (“as Naval said…“)
  • “Hot take” without actual depth

Score:

  • ✓ — clip contains a reframe of conventional wisdom OR a sharp observation that lands as “huh, true”
  • ⚠ — perceptive in spots but mostly received
  • ✗ — clip contains zero original observation; viewer leaves no smarter

7. Illuminated

What it is: clip ends and the viewer sees something more clearly than before. Vision-bearing close. Promise-anchored. Not a sells-y CTA.

What it’s NOT:

  • “Link in bio” / “preorder now” / “buy this”
  • “Hope this helped!” — completion-feeling, no illumination
  • Open-loop without payoff
  • Trailing-off close

Score:

  • ✓ — clip ends with a promise-close OR a vision sentence that recontextualizes the opener
  • ⚠ — ends with a quiet observation (acceptable if earned)
  • ✗ — sells-y close OR fade-out OR random tail

Anti-patterns the sub-agent must detect

Specific delivery / phrasing patterns that violate the register. Each gets flagged with an explicit name in virality-pass.md so Daniel sees what’s wrong.

Motivational-speaker patterns (BANNED)

  • Rising-intonation lift-statement: “And THAT is what changed everything.” → declarative-as-question rhythm. Detect via terminal pitch rise + emphasis on penultimate word.
  • List-of-three rhetorical scaffold (when not earned): “It’s about X, Y, and Z.” → tricolon as performance device. Allowed if X/Y/Z are all specific and earned; flagged if abstract.
  • “And here’s the kicker…” performative pivot: any phrase that BUILDS UP to a payoff. The payoff should be felt, not announced.
  • “Trust me on this one” / “I promise you” — meta-assertions that try to manufacture credibility.
  • “Most people don’t realize…” — false-contrarian opener.

Hustle-bro patterns (BANNED)

  • Performance-vocabulary: “grind / hustle / 4am / sigma / built different / on a Tuesday” — entire vocabulary excluded.
  • Outcome-flexing: “$1M ARR in 90 days” without genuine context → bait. Acceptable WITH context that explains the trade-off.
  • Solution-positioning: “if you’re not doing X, you’re missing out” — DTC-pitch energy.
  • Audience-condescension: “let me drop some knowledge” / “real ones know” / etc.

Productivity-influencer patterns (BANNED)

  • Routine-prescription: “here’s my morning routine” / “my 3-step framework” → unless tied to a specific build moment, not generic advice.
  • Tool-stack flexing: mentioning tools to seem knowledgeable, NOT in service of one job. (Distinguishes from Pillar 7’s stack-naming, which IS the substance.)
  • Optimization-fetish: “I cut my X by Y%” with no human stake.

TED-pacing patterns (BANNED)

  • Rhetorical question opener: “Have you ever wondered…?” → buys time, doesn’t pull.
  • Build-up to “the truth”: TED-style escalation toward a thesis 60s in. A Friend opens on the thesis.
  • Story-then-point: narrative anecdote that exists ONLY to deliver a generalized lesson. Daniel’s vulnerability beats are stories WITH points; they’re not stories that EARN points.

Brain-rot patterns (BANNED — separate gate)

  • Engagement-bait: “you won’t believe what happened next” / “wait for it…” with weak payoff.
  • Manufactured drama: raised stakes that aren’t real (“I almost didn’t post this” when posting is fine).
  • Specific-number for shock: “1,247 hours of work” without stake. Number must pair with substance (Pattern A from shortform-virality.md).
  • Performed vulnerability: “Today I want to share something deeply personal…” preamble.

Detection heuristics (what the sub-agent computes)

The sub-agent doesn’t have a tone-classifier. It uses surface-text proxies:

Anti-patternDetection signal
Motivational lift-statementLast 3 words bold-emphasized in transcript context, exclamation-energy adjective (“AMAZING”, “INSANE”)
Hustle-bro vocabularyWord-list scan: grind, hustle, sigma, built, 4am
Performance-vocabularyWord-list scan: revolutionary, seamless, leverage, game-changer, AI-powered
Rhetorical question openerFirst sentence ends in ”?” AND second sentence starts with “Well…” or “The truth is…”
TED build-upOpener uses second person + question pattern + delayed thesis (>10s before declarative)
List-of-three (unearned)“X, Y, and Z” structure where X/Y/Z are abstract nouns
Brain-rot baitPhrases: “wait for it”, “you won’t believe”, “the result”, “this is wild”
Manufactured dramaStake doesn’t have specific number / date / named-thing AND uses “almost” / “nearly” / “could have”

These heuristics are coarse — they catch patterns, not intent. The sub-agent surfaces detected matches in virality-pass.md with the named anti-pattern. Daniel decides if it’s a real violation or a false positive.

What “anti-slick” means in execution

When the sub-agent recommends a cut or reorder, it must NOT optimize for slickness. Specifically:

  • Don’t smooth the cadence. If Daniel’s natural sentence has a rough edge, leave it. Smoothing reads as scripted.
  • Don’t optimize for “clean” delivery. Restart-pauses, breath, mid-sentence corrections — these earn realness. Cut filler (“uh”, restarts that go nowhere) but keep cognitive effort visible.
  • Don’t punch up the language. Don’t propose synonyms that “land harder.” Daniel’s word IS the right word — verbatim contract is hard rule.
  • Don’t add transitions. No “but here’s the thing” inserts, no “let me explain” bridges. The cut between two beats IS the transition.
  • Don’t trim toward virality if it costs realness. A clip that hits engagement-metric perfection but loses Daniel’s voice fails the “feels like me” gate. Voice wins.

This is the line between engagement-aware (allowed) and engagement-bait (banned). The seven-axis rubric is the test.

What the sub-agent does with this canon

  1. Scores every draft clip against the 7 axes. Each gets ✓/⚠/✗.
  2. Detects anti-patterns via heuristics. Names the pattern in virality-pass.md if hit.
  3. Refuses to propose changes that violate the register. Even if a change would boost retention.
  4. Surfaces register-drift warnings. Two ⚠ across the 7 axes = “register drift” flag (does NOT block — Daniel decides).
  5. Hard-blocks on banned anti-patterns. Brain-rot patterns / hustle-bro vocab / motivational-speaker rhythms = BRAND-NONNEGOTIABLE-VIOLATION tag, blocks output until cleared.

Sources & canon basis

  • _brand/voice-preservation.md §5 — banned vocabulary list (revolutionize, seamless, AI-powered, game-changer, leverage, exclamation points).
  • _brand/voice-preservation.md §3 — two-beat cadence patterns.
  • _brand/voice-preservation.md §6 — tool-naming rule (Pillar 7).
  • _brand/pillars.md — 7 pillars + anti-patterns (per-pillar).
  • references/attention-preservation.md “Anti-Brain-Rot constraint” — engagement-aware vs engagement-bait line.
  • references/essence-extraction-framework.md — Schärfung/Trim/Essenz; this file is the upstream filter.
  • A Friend brand book (A Friend Design System/SKILL.md) — voice section. Source-of-truth for any conflict.
  • Daniel verbatim 2026-05-03 — the seven adjectives are the authoritative list. Sub-agent does not extend or substitute.

When the sub-agent flags an anti-pattern, it cites THIS file or one of these sources. Not generic creator advice.

Graph View