Structure non-fiction content as a sequence of curiosity loops with varied rhythm, conversational tone, written ending-first, on a non-obvious lens, with a visual hook. Use when drafting or rewriting blog posts, video scripts, social posts, founder essays, or pitch narratives where structural craft (not idea generation) is the bottleneck.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Structure non-fiction content as a sequence of curiosity loops with varied rhythm, conversational tone, written ending-first, on a non-obvious lens, with a visual hook. Use when drafting or rewriting blog posts, video scripts, social posts, founder essays, or pitch narratives where structural craft (not idea generation) is the bottleneck.
Use this skill to make a non-fiction piece keep a reader past the first 5%, the first 30 seconds, the first 200 words. The traditional Freytag bell curve was built for captive audiences. On the internet there is no captivity — viewers re-decide whether to stay every few seconds. A story-craft agent's job is to detect where the climb breaks and rewrite that specific rung, not to "improve the writing."
A story works when six craft moves (dance, rhythm, tone, direction, lens, hook) pace a viewer up a six-rung dopamine ladder (stimulation → captivation → anticipation → validation → affection → revelation), with seven specific failure modes filtered out at the QA pass.
When to Use
Draft or rewrite a blog post, essay, video script, or social post that already has a thesis but is structurally flat.
Audit a draft that "feels off" — diagnose which rung of the ladder fails first, then rewrite that rung.
Convert a brain-dump or transcript into a structured piece.
Stress-test a founder narrative, pitch deck, or fundraising story for retention craft.
Generate candidate lenses when a topic is saturated and the founder needs a non-obvious angle.
Build a long-form piece by stacking 3–7 nested curiosity loops instead of one long arc.
Do not use this skill for cold idea generation, voice/persona development, or the upstream "what to write about" decision. For that, pair with nonfiction-writing-from-lived-conviction (lived experience → inquiry) and content-strategy-beyond-blogging (intent → format). Use this skill on the structural layer between idea and ship.
Core Principles
The viewer is not captive. Every paragraph or 5-minute window is a re-decision to stay. Earn it; don't assume it.
Open loops before you close them. Conflict opens loops. Context closes them. Stories perform when loops keep opening before they close.
Rhythm is structural, not stylistic. Uniform sentence length is what makes viewers swipe. Varied sentence length is what makes prose sing.
Write the ending first. You cannot construct the dance until you know where it lands.
The lens beats the topic. Anyone can cover a topic. Only one founder can cover it through their lived experience.
Show before you tell. Visual hooks beat audio-only hooks ~10×. For text, that means concrete image / number / scene before the abstraction.
Stack rebuys, don't ride single peaks. Solve a problem; immediately open a new one. After 3–4 cycles, the viewer feels you solved 4 problems but you really solved 1.
Diagnose by ladder, not by feel. When a piece flops, walk the six rungs in order, name the first failure, fix that one specifically.
Craft requires reps. Knowing the rules is easy. Applying them consistently is what separates 10× creators. Draft 1 is a diagnostic surface, not a ship bar.
The Three Pillars
Pillar 1: The Dopamine Ladder (the psychology spine)
Six rungs the viewer climbs in order. Each rung releases more dopamine than the last. Skipping a rung collapses the climb. Rungs 1–4 are scriptable per piece. Rungs 5–6 accumulate across many pieces and bind a viewer to the messenger.
Rung
Window
Trigger
Dopamine size
Failure mode
Stimulation
0–2s pre-cog
Color, motion, contrast, brightness vs. surrounding feed
Smallest, mandatory
Looks like everything else; eye doesn't lock
Captivation
2–7s
Open question implants in viewer's head
Largest in-loop until validation
No question pops, or pops but irrelevant
Anticipation
7s → loop close
Controlled fact leak; viewer guesses answer in real time
Peaks just before validation
Confusing facts; viewer can't anticipate
Validation
Loop close
Non-obvious resolution — twist (entertainment) or counter-intuitive insight (education)
Substantial — only if non-obvious
Predictable payoff or cliffhanger
Affection
Cross-piece
Recognisable face/voice + likability levers
Persistent — buys leash on future pieces
Faceless, inconsistent, performative
Revelation
Cross-channel
Repeated non-obvious value in a defined problem area
Maximum — fires on sight of name
Topical drift; nice-to-have niche
Curiosity dopamine fires on the question, not the answer. Anticipation dopamine peaks just before the answer. Validation only releases dopamine if the payoff is non-obvious — predictable closes still count as a watch but produce little reward and don't drive return.
Pillar 2: The Six Craft Moves (the construction blueprint)
Run these as ordered passes when constructing or rewriting a piece.
Dance — alternate context and conflict using "but / therefore," never "and then." Outline the piece as discrete beats. Between every two beats, force the insertion of but or therefore as a connector. Reject and then. Aim for 3–4 conflict loops in the first 30 seconds of short-form, 2–3 in the lede of a blog post. (Stone & Parker via Kallaway.)
Rhythm — vary sentence length so the script sings. Render the draft one sentence per line. Inspect the right-hand edge: a straight edge is dead rhythm. A jagged edge confirms varied rhythm. Mix short, medium, and one long crescendo sentence per major section. Read aloud (TTS) to surface monotony. (Gary Provost via Kallaway.)
Tone — write to one named close friend, not "to camera." Ask the founder to name the one person they're writing to. Persist the named reader across drafts. Reject broadcast phrasings (In this article we will explore…) and rewrite as direct address (You probably know the feeling…). (Steve Jobs / Casey Neistat / Emma Chamberlain via Kallaway.)
Direction — write the last line first. Force the ending before the body. Evaluate it against a memorability bar: would a stranger text this to a friend? For looping content, test whether the last line + first line read as a continuous loop. Treat last-line-first as a hard gate, not a polish step. (Christopher Nolan via Kallaway.)
Lens — pick a non-obvious angle on the same topic. For any topic, generate at minimum 5 candidate lenses ranked by obviousness. Default to the lens where the founder can be category-of-one. Refuse Tier 1 (saturated, descriptive) lenses unless a non-obvious proof is attached. Source the strongest lens from the founder's lived operating experience — pair with nonfiction-writing-from-lived-conviction. (Taylor Swift via Kallaway.)
Hook — punchy first line that names the topic, paired with a visual that shows it. First line names the topic in 6–8 words. Reject opaque openers (Wait till you see this, You won't believe this). For video, pair every hook with a visual cue that confirms the topic before the audio finishes. For blogs, the analog is a vivid concrete image, specific number, or verbatim quote in the first sentence — never a throat-clearing intro paragraph. (Epic Gardening via Kallaway.) For full hook craft, defer to hook-craft-short-form.
Lens stratification (Tier ladder)
Tier 1 (saturated): Surface-level descriptive coverage everyone is doing. Reject by default.
Tier 4 (category-of-one): Lens only this founder, with their lived experience, can credibly hold.
Default target: Tier 3 or Tier 4.
Pillar 3: The Seven Mistakes (the QA reject pass)
Run these as a final filter before ship. If any mistake fires, route back to the corresponding craft move.
The Traditional Story Arc. Bell-curve buildup, single climax, late stakes. Loses 80% before rising action. → Replace with the Kallaway Story Arc: start at ~70/100 intensity, spike to ~90 within 1–2 minutes, release to ~30, re-peak to ~75 every 2–5 minutes.
Jumbling The W's. Leading with where/when ("It was a cold December night…"). Wastes the 30-second window. → Restack as what + why → who + how → where + when. Lead with the cake; demote the icing to position 3+.
Single Hook, No Rebuy. One peak then coast. Average view duration craters; algorithm punishes. → After every resolution, open a new loop within 30 seconds with bridge phrases like "but the truth is…", "the catch is…", "even with that solved, you still face…"
Missing Villain. Hero alone with no antagonist. No tension, no stakes. → Pick a thing/person/group/methodology/status quo/incumbent as the antagonist. Frame contrast as they do X, but we do Y.
Nobody To Root For. Stakes for the creator but not the audience. Viewer is a spectator without a team. → Insert explicit audience-mapping ("for entrepreneurs," "if you're a solo founder," "for people who…"). Name the viewer's situation; tie their stakes to the hero's.
Lacking Atomic Shareability. Story can't be compressed to a Paul-Revere-tier sentence. → Generate a 1-sentence atomic version (≤10 words). For complex points, double-encode: explain once for the colleague, restate once as a metaphor for the child.
Not Painting The Picture. Words alone, no visuals. Brain comprehension drops without visual anchors. → Annotate every line with a "needs visual?" tag. Generate b-roll briefs alongside the script. For abstract points without planned visuals, propose a metaphor or analogy that can be drawn.
Workflow: Generate A Piece From Scratch
Given a thesis, target reader, and content type:
Source the inquiry first. If the founder hasn't done the upstream work, defer to nonfiction-writing-from-lived-conviction to extract the lived material, contradiction, or hard-won opinion this piece will explore.
Generate 5+ lenses. Stratify by tier. Default to Tier 3 or Tier 4. Reject Tier 1 unless a non-obvious proof is attached.
Write the last line. "The last dab" — so memorable a stranger would text it to a friend. For looping content, write the first line at the same time and test continuity.
Outline beats. Discrete beats, not paragraphs. Between every two beats, insert but or therefore. Reject and then. Map the intensity curve: open at ~70/100, hit ~90 within the first 1–2 minutes (or first 200 words), release to ~30, re-peak every 2–5 minutes (or every 400–600 words).
Restack the W's. What + why first. Who + how second. Where + when last.
Stack rebuys. For each peak-and-release, open a new loop within 30 seconds (video) or one paragraph (blog). Aim for 3–4 closed-loop cycles across the piece.
Name the one reader. The founder writes to that one person, by name, in their head. Reread as if texting them.
Run the ladder check. State the question implanted at the 5-second mark (or first paragraph). State what the viewer is anticipating mid-piece. Verify the validation is non-obvious.
Pair every point with a visual. B-roll brief per beat for video. Hero image / scene / number / quote for blog. For abstract points without a visual, swap in a metaphor that can be drawn.
Draft prose using rhythm rules. One sentence per line. Mix short / medium / long. Read aloud. Confirm the right-edge jaggedness.
Write the hook last. Once the body and last line exist, write the opener. First line names the topic in 6–8 words. Defer to hook-craft-short-form for the full slot grammar / archetype / four-mistake diagnostic on the hook itself.
Run the seven-mistake reject pass. Walk the seven mistakes in order. On any fire, route back to the corresponding craft move.
Workflow: Audit And Rewrite An Existing Draft
Given a draft the founder already wrote:
Walk the dopamine ladder in order. State explicitly:
Did the first 1–2 seconds (or first sentence) visually distinguish from the feed / noise?
At the 5-second mark (or first paragraph), what is the question implanted in the viewer's head? If you can't articulate it, captivation failed — fix that first.
Mid-piece, what is the viewer anticipating? If unclear, anticipation broke.
Is the payoff non-obvious? Could a smart viewer have written it before consuming?
Across the channel, is the persona / problem area consistent?
Stop at the first failed rung. Do not propose holistic rewrites. Fix that rung; re-evaluate.
Run the seven-mistake reject pass. If a mistake fires, route to the corresponding craft move.
Rebuild the prose only after structure is clean. Apply Rhythm and Tone passes last; they are surface-level relative to the ladder rungs and craft moves.
Long-Form vs Short-Form Translation
Kane's mechanics are short-form-pilled. Translation table for blog posts and long-form essays:
Short-form rule
Blog / long-form analog
4 loops in first 30s
2–3 loops in the lede (first 200 words)
Visual hook in first second
Concrete image, scene, number, or verbatim quote in first sentence
Last-line "last dab"
Final-paragraph kicker that survives in isolation
Looping last-to-first
Final paragraph that recasts the opening claim with new evidence
Right-edge jaggedness check
Same — works in any rendered prose
Re-peak every 2–5 minutes
Re-peak every 400–600 words
Visual b-roll per beat
Pull-quote, sub-heading, image, code block, or table per major beat
Talking-head + captions only
Wall-of-text without sub-heads, images, or pull quotes
Loops within long-form should be nested (3–7 mini-loops inside the larger arc), not stretched into a single long anticipation curve.
BuildOS Voice Translation
The Kallaway register skews toward shock and creator-economy hype ("0 to 100K subs," "the algorithm hates you"). BuildOS positions anti-AI, anti-feed, anti-hype. The mechanics hold; the register changes.
Default villain set for BuildOS pieces: AI-everywhere autopilot, the productivity-app graveyard, context-switching, task-list theater, social-as-feed (vs. interest media), founder loneliness sold as hustle.
Default rooting interest: the messy-thinker, the ADHD operator, the founder building complex things, the person making something who hates how AI tools sound. Audience-map explicitly in the opening.
Lens default = Tier 4 (category-of-one) sourced from the founder's actual operating experience. The conviction skill upstream is mandatory, not optional, for BuildOS philosophy posts.
Validation register check: payoffs in BuildOS pieces should land as relief (the contradiction resolves into honest action) rather than shock (the viewer feels duped into agreeing). If the only way the payoff lands is "haha gotcha," reject it.
No manufactured rebuy stakes. Each new loop must be a real stake the body can defend, not a manufactured cliffhanger to keep AVD up. Manipulative rebuys break the brand.
Don't perform "founder vulnerability." Tone pass should produce direct address to one named reader, not confessional posture for engagement.
Guardrails
No and then connectors between beats. Reject any outline where every transition is additive.
No bell-curve intros. Pieces opening below ~70/100 intensity (scene-setting, throat-clearing, "for years I've wondered…") fail by default.
No leading with where/when. What + why first.
No single-loop pieces. A piece with one peak and a slow descent will lose 80% before the descent finishes.
No hero without a villain. If you can't name the antagonist (process, methodology, incumbent, status quo, idea), kill the draft.
No piece without explicit audience-mapping in the first 30 seconds (video) or first paragraph (blog).
No "knowing-only" payoffs. Predictable validation produces no dopamine. If a smart on-topic reader could have written your payoff, rewrite or layer in a counter-intuitive frame.
No talking-head-only video longer than 60 seconds without at least one visual cue.
No prose without rhythm variation. Standard-deviation of sentence length below threshold = monotony = swipe.
No piece without a last line written first. No exceptions; this is upstream of the body.
No Tier 1 lens unless a non-obvious proof is attached.
No level-5/6 advice ("be more likable") for creators who haven't fixed levels 1–4. And no level-1/2 advice ("strengthen your hook") for creators with traction but no fandom — they need consistency, not a better hook.
No personality forcing. Recommend "increase concreteness of the problem solved + visible passion + reduce filler" — never "be more charismatic."
No drift from the channel's identity contract. Every shipped piece reinforces the founder's defined problem area; flagged drift erodes Rung 6 (revelation).
No manipulative rebuy stacking. Each new loop must be a real stake. Manufactured cliffhangers break trust faster than they extend AVD.
Output
Return a structured piece bundle:
Inquiry / thesis — sourced from nonfiction-writing-from-lived-conviction upstream
Lens — chosen tier, why category-of-one for this founder, 3–4 alternates considered
Last line — written first, memorability test passed
First line — paired with the visual / image / scene / number / quote
Beat outline — discrete beats, every transition tagged but or therefore
Intensity curve — opening ≥70/100; first peak within 1–2 min or 200 words; re-peaks every 2–5 min or 400–600 words
W-stack — what + why → who + how → where + when sequence in opening
Rebuy map — each closed loop paired with the next loop opening
Audience map — explicit "for X" statement and tie-in to viewer's stakes
Villain — named antagonist (entity, process, methodology, status quo)
Atomic shareability line — one Paul-Revere sentence, ≤10 words
Visual brief per beat — what the reader/viewer sees during each
Tone audit — named reader, broadcast phrasings stripped, direct address verified
Ladder diagnostic — implanted question at 5s/first-paragraph; mid-piece anticipation; non-obviousness rating on payoff
Seven-mistake reject pass — each mistake marked clean or fail-with-fix
Identity-contract check — does this piece reinforce the founder's defined problem area? If no, flag for drift
For audit-mode runs, replace the first set with a diagnostic report: the first failed rung, the corresponding craft-move fix, and the rewrite plan.
Source Attribution
Distilled from three Kallaway videos plus the existing BuildOS conviction skill:
How To Become A Master Storyteller — six craft moves (dance, rhythm, tone, direction, lens, hook), Stone-Parker but/therefore rule, Gary Provost rhythm passage, "last dab" / write-the-ending-first, lens stratification (Taylor Swift), "show while you tell."
How to Become a Storytelling Genius (Dopamine Ladders) — six-rung dopamine ladder (stimulation / captivation / anticipation / validation / affection / revelation), pre-cognitive 200ms processing, curiosity-fires-on-the-question, peak-just-before-the-answer, non-obvious-payoff requirement, video-level vs channel-level distinction, likability levers, the "edging of the storytelling world."
7 Storytelling Mistakes That Are KILLING Small Creators — Kallaway story arc (replaces Freytag), W-stack reorder, rebuy/rehook mechanic, villain framing, audience-mapping for rooting interest, atomic shareability / Paul Revere test, double-encode tactic, "paint the picture."
Pairs upstream with nonfiction-writing-from-lived-conviction (lived experience → inquiry → contradiction) and downstream with hook-craft-short-form (final hook polish using slot grammar, three-beat structure, six archetypes, four-mistake diagnostic). Cross-link with content-strategy-beyond-blogging for the intent-to-format decision above the structural craft layer.
