Plan and audit publishing decisions on interest-media platforms with full awareness of the matching loop, viewer psychology, and BuildOS's anti-feed brand stance. Use when picking a hero platform, narrowing a topic, scoring whether to publish a specific piece, deciding what to keep / modify / reject from creator-economy tactics, or running a dual audit (algorithm-clean AND brand-coherent) on a draft.
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.
Plan and audit publishing decisions on interest-media platforms with full awareness of the matching loop, viewer psychology, and BuildOS's anti-feed brand stance. Use when picking a hero platform, narrowing a topic, scoring whether to publish a specific piece, deciding what to keep / modify / reject from creator-economy tactics, or running a dual audit (algorithm-clean AND brand-coherent) on a draft.
Use this skill to publish into interest-media platforms without becoming captured by them. The platforms are matching engines, not megaphones — every video is fingerprinted multimodally, tested on a sample of ~200 mostly-non-followers, and either boosted, retried, or killed based on the response. The creator's only real leverage is to make matching easy (narrow topic + consistent avatar) and to make the sample stay (genuinely useful content for that avatar).
The hard part for BuildOS is that the same playbook that wins distribution also includes tactics that violate the brand. This skill encodes the dual audit: every published artifact must be (a) clean on the matching loop AND (b) routing attention toward the BuildOS interest-media diagnosis, not away from it. Algorithm-aware without being algorithm-captured.
When to Use
Pick a hero platform for a six-month focus window.
Narrow or defend a topic / avatar pairing across many posts.
Score whether to publish a specific piece — does it strengthen or pollute the matching loop?
Decide what to keep, modify, or reject from creator-economy tactics (manufactured FOMO, comment bait, urgency stacks, re-stoked pain at close).
Audit a published piece that flopped — diagnose where in the loop it broke (topic mapping / sample group / engagement signals).
Run the dual audit (Kane mechanics + BuildOS thesis) on a draft before ship.
Plan rented → owned ramps (feed → newsletter → product) as the actual brand strategy, not a side benefit.
Build a "winners library" of outlier videos in the niche before brainstorming.
Do not use this skill for hook craft (defer to hook-craft-short-form), narrative shape (defer to story-driven-content-craft), script structure (defer to viral-video-script-structure), or upstream "what to publish" thesis selection (defer to nonfiction-writing-from-lived-conviction and content-strategy-beyond-blogging). This skill is the distribution and brand-coherence layer above all of those.
Core Principles
Social media is no longer social. It is media. The 2020 TikTok-driven shift collapsed the social graph as a primary distribution input. Every major feed runs on audience matching, regardless of follow relationship.
The platform's only goal is session length. Distribution decisions serve one objective: keep the user on the app longer so more ads run. Reframe "will the algorithm like this?" → "will the next viewer stay on the platform longer because of this?"
Audience matching is the highest-leverage lever. Narrow topic + single avatar consistently across many posts. Mixing topics produces a blended fit score and poisons the next several videos.
On-target virality beats pure virality. A 200K-view hit inside the avatar is a better business outcome than a 10M-view hit that spans disjoint groups — the broad hit corrupts your next-video targeting.
Treat platforms as islands. Cross-platform forced ramps nuke engagement. Build world-cohesion across islands; don't force discovery between them. The only sanctioned ramp is rented → owned.
Value does not accrue at the media layer. AdSense and brand deals will not produce real money. Use media to manufacture attention; route attention to owned offerings (BuildOS itself, in this case).
Honest signals over manufactured signals. Watch time and completion are aligned with quality; comment bait, manufactured FOMO, and faux credibility are not. The brand cannot survive feed-style manipulation even when feed-style manipulation works.
The dual audit is non-negotiable. Algorithm-clean is necessary. BuildOS-thesis-coherent is also necessary. Either one alone is failure.
Stop tweaking the wrong layer. Posting time, hashtags, captions don't matter. Topic precision, hook quality, and the four engagement attributes do.
The Four Pillars
Run them in this order. Pillar 0 is the strategic decision above the matching loop — skip it and you optimise the wrong system.
Pillar 0: Pick Your Content Game (strategic decision before any tactics)
Five different content games run on social media simultaneously. Each has different rules, different objectives, different ways to win. The #1 reason businesses waste months on content is using the wrong game's playbook. Pure virality is correct for Games 1 and 2. For Games 3, 4, 5, an off-target viral hit poisons the next several videos.
Game
Name
Examples
Goal
Monetization
1
Mainstream Fame
Mr. Beast, Charli D'Amelio
Max total views
Ad revenue + broad brand deals (media layer)
2
Category Fame
Dad Social, Cleo Abram
Max views in a niche
Niche brand deals + sponsorships (media layer)
3
Intuitive Products
Clothing brands, CPG
Build a world around product
Vibes / lifestyle-fit purchases
4
Explanatory Products
AG1, Meta Ray-Ban glasses
Educate the buyer
Education-driven purchases
5
Authority Building
Consultants, coaches, info products
Build deep trust + expertise
Off-platform conversion (leads / clients)
Quick diagnostic:
If you sell…
Play game…
Services, coaching, consulting, info products
Game #5
Commodity consumer product
Game #3
Differentiated product needing explanation before purchase
Game #4
No offer yet but want a targeted audience
Start #2 → shift to #5
The Trust Formula (Game #5 specifically):
Trust = Minutes Watched × Relevance × Usefulness
Multiplicative. Zero on any axis = zero trust accumulated. A long video with low usefulness builds nothing. A useful video to the wrong audience builds nothing. 1,000 high-trust views from potential buyers is worth more than 100,000 disconnected views from the wrong audience.
For Game #5 the unit of measurement is trust accumulation, not views. View count is vanity for off-platform-monetizing creators. Cohort-level trust signals (returning-viewer minutes, comment specificity, off-platform conversion rate) are the real channel-health metric.
The Off-Target Cost Heuristic: when evaluating an idea that would go viral but to the wrong audience —
Estimated upside = (off-target views × ~0% conversion) ≈ 0 leads
Estimated cost = (next 5–10 videos × suppressed reach) = N suppressed leads
Decision: refuse the off-target idea unless the algorithmic cost is explicitly justified.
Kallaway's own confession: his Ironman video hit 46M views in 5 days and produced zero customers. Ego full, bank account empty. Going viral with the wrong audience is worse than not going viral at all.
BuildOS classification: Game #4 (Explanatory Product) wrapped in a Game #5 (Founder Authority) shell. The two games complement at the channel level but cannot mix inside a single piece:
Game #5 evidence → anti-feed essays, lived-experience POV pieces, contrarian takes on AI-everywhere, founder-direct philosophy. These prime trust upstream.
Game #4 evidence → BuildOS demos, brain-dump-to-structure walkthroughs, feature explainers, comparison videos. These convert that trust into sign-ups downstream.
Channel-level rule of thumb: ~3:1 ratio of Game #5 (authority) to Game #4 (product). Tag every shipped piece with its game in the content log so the ratio can be audited. Each piece picks one game. Mixed-game pieces get split or rejected.
Pillar 1: The Matching Loop (mental model)
Topic discipline (transcript + caption) →
Algorithm classifies the topic via multimodal fingerprint
(computer vision + audio/transcript + metadata) →
Algorithm picks ~200 mostly-non-follower viewers it predicts will enjoy →
Sample group engages (or doesn't) on three signals:
- Average watch time + percent completion
- Engagement rate (likes + comments + shares / views)
- Watch-time session share (silent boss metric)
→
Positive data → push to ~2K, then ~20K, then ~200K (tier shifts)
Neutral data → re-sample another ~200; do not scale up
Negative data → tighten and stop pushing
Every distribution decision flows from understanding this loop. Two creator levers exist:
Better fit score. Narrow topic + single avatar + multimodal signal coherence (visuals, transcript, metadata all reinforce the same topic).
Sample group engages. Driven by the four-attribute engagement filter (see Pillar 2).
What does not matter: posting time, hashtag stuffing, caption text beyond minor metadata weight, "settings hacks." Treat advice that leads with these as a credibility flag against the source.
The platforms are also islands, not an ecosystem. Posting URLs in tweets, LinkedIn posts, or IG Stories nukes engagement; users consume at platform-specific speeds (TikTok ~50mph, IG ~35, YouTube ~10). Native execution per island; world-building (visual / verbal / aesthetic cohesion) for cross-island recognition; rented → owned for actual conversion.
Pillar 2: The Quality Gate (four-attribute filter + Trust Formula + six checkpoints)
The cleanest BuildOS-aligned content quality gate. Every piece must hit all four:
Topic relevance — solves a problem the avatar actually has.
Non-obvious AND tactically implementable — new to them and usable.
High absorption — the viewer can follow the explanation.
Short distance to implement — small action, big result.
These four attributes produce engagement signals as a byproduct of quality, rather than tricks that simulate engagement. They double as commercial DNA — the same four turn viewers into buyers.
For Game #5 pieces specifically, sharpen the four-attribute filter into the Trust Formula — multiplicative scoring on a 1–5 scale per axis:
Axis
1
3
5
Minutes Watched
<30s avg
~50% completion
Re-watches + binge-throughs
Relevance
Wrong audience
Adjacent audience
Direct buyer avatar
Usefulness
Inspirational only
Theoretical insight
Walk-away tactic that produces result
Multiply the three. Below ~30/125 → the piece is leaking trust, not building it. Don't ship.
Layered underneath, every value-driven piece walks the viewer through six checkpoints in sequence. Any failed checkpoint = bounce. Fix in order; don't optimize gate 5 if gate 1 is broken.
Gate
Question
What's at stake
1
Painpoint Acceptance — do they admit the pain?
Hook
2
Sprint to Trust — do they believe you?
First 10–30 seconds
3
Plan of Attack — different, structured path?
Frame
4
Likeability — do they want YOU to deliver?
Voice / persona
5
Harness Attention — do they keep watching?
Body / rehooks / rhythm
6
Drive Action — do they do the thing?
Outro / CTA
The Trust Hierarchy at gate 2 (most → least effective):
Testimonial of solving this for someone like the viewer.
Proof you solved it for yourself.
Massive following or trusted co-sign.
Evidence you've crossed the same gap.
Common ground / shared identity.
Studio quality / professional setup.
Top 3: any one is sufficient. Bottom 3: may need multiple stacked. Bigger ask = more trust required. Deeper pain = less trust required.
Pillar 3: The Dual Audit (Kane mechanics + BuildOS thesis)
The central skill of this combo. Every piece runs through both audits.
Kane audit (algorithm-clean):
Does the visual / transcript / metadata reinforce the same topic?
Does it match the avatar this channel has trained the algorithm on?
Is the opening designed for the non-follower sample, not for existing followers?
Does it pass the four-attribute engagement filter?
Is there proof, or evidence-of-shared-gap, or common ground at gate 2?
Does the body have a Plan of Attack with a real point of difference, not surface contrarianism?
BuildOS audit (brand-coherent):
Does this piece route attention toward the BuildOS thesis (anti-feed, anti-AI-as-autopilot, thinking environment for people making complex things) — or away from it?
Does it execute the matching loop without manufacturing synthetic volume?
Is the contrarian take from real conviction the founder would defend even if comments were disabled?
Is the persuasion stack free of manufactured FOMO, manufactured urgency, faux credibility (polish-over-proof), and re-stoked pain at close?
Does the rented → owned ramp lead to BuildOS itself (the offering layer), not to AdSense / brand deals (the media layer)?
Would this piece still be worth publishing if every metric were invisible?
A draft must pass both audits to ship. Algorithm-clean alone optimizes the matching loop and treats the feed as neutral terrain; the founder's voice flattens. BuildOS-coherent alone produces work that doesn't get distributed. The skill enforces both.
The Keep / Modify / Reject Catalog
Tactics from Kane's "18 Psychology Tricks," tagged for BuildOS use. Use this as the operating manual.
Keep (12 — use directly)
#
Tactic
Why kept
1
One Standard Deviation Pain
Honest reframing — adjacent to the real desire so the viewer makes the leap. Already our anti-AI strategy.
3
Audience of One
Plain second-person writing craft. Audit you vs I ratio in the first 30s.
4
Painpoint–Solution Gap
Honest when both states are real. Map "messy thinking → structured work."
5
Proof
Real proof artifacts, ordered by relevance to the specific viewer. No fabrication.
6
Common Ground
Anchor in specific shared experience, not generic "I've been there." Lean on DJ's actual lived ADHD/founder context.
8
Point of Difference
This is the entire BuildOS positioning. Common belief stated, contrarian-but-true alternative held with conviction.
9
Ordered List
Cognitive scaffolding, not manipulation. Use meta-grouping when the headline number is large (18 → 6 checkpoints).
11
Simplification
Plain language non-negotiable. Pair jargon with metaphor.
12
Authenticity
Same person on/off camera. The entire BuildOS founder-voice doctrine.
15
Trance Rhythm
Vary sentence length. Jagged right-edge in the script doc. Prose craft, full stop.
17
Unexpected Surprise
Over-deliver. Free guide content above paid course quality. Aligns with "give the playbook away."
18
The Recap
Honest closure. Recap of points, pair with one specific CTA.
Production quality high but human (DJ on a walk, room-tone recordings). Never let polish carry weight that proof should carry. Too polished signals "ad" and breaks anti-feed authenticity.
10
Energy
Don't let energy decay across paragraphs. But MrBeast pace breaks the brand. Our energy is intensity of insight, not volume.
13
Create a Hunt
Front-load the best stuff (respect for viewer's time) — yes. "Wait for the end, the last one is killer" gimmick — only with a real long-loop payoff, never to inflate watch time on weak content.
14
Rehooking
Re-orient to real stakes — yes. "If you don't figure out this next one, you lose forever" escalation — no. Use re-hooks to remind, not to threaten.
Reject (4 — out of scope for BuildOS publishing)
The manipulation core: any lever that manufactures a reaction the content hasn't earned. All four corrupt the next
piece's targeting even when they win the current one's metrics.
#
Tactic
Source
Why rejected
16
Close the Loop — re-stoke pain at close to drive paid action ("but if you want speed, buy this")
18 Psychology Tricks
Re-stoking pain immediately after claiming you've solved it is the manipulative core. The honest version: "Here's everything I'd do. If BuildOS helps, here it is — try free." Trust the viewer to choose.
SA-1
Manufactured contrarianism / comment-bait — "pick the contrarian side," ratchet up stance language, "create more enemies"
Social Algorithms
Extracts comments by staging a fight the creator doesn't believe in. Real stances from real conviction generate real engagement; the comment-bait ladder trains the algorithm on the wrong audience.
SA-2
Rage-bait / fear / drama as engagement levers
Social Algorithms
Weaponised negative emotion travels but poisons the brand and the matching signal. The anti-feed positioning cannot survive feed-style manipulation even when it works.
SA-3
Audience-poaching — build topics around cult-loved brands to siphon their audience
Social Algorithms
Borrowed audiences don't match the avatar; the spillover corrupts the next several pieces' targeting the same way an off-target viral hit does.
Take real stances from real conviction; skip the comment-bait ladder. (Reject rows tagged SA-* come from Kane's
"Social Algorithms" comment-driver stack; #16 from "18 Psychology Tricks.")
Workflow: Plan A Publishing Strategy
For a new channel, account, or six-month focus window:
Pick the content game (or game ratio). Run the Pillar 0 diagnostic. For BuildOS: confirm the Game #4 + Game #5 hybrid and set the ratio (default ~3:1 #5:#4). Tag every piece with its game in the content log.
Pick the hero platform. Where do existing customers cluster? Where does the founder learn the most about the space? Pick one + email. Six months. No exceptions.
Define the topic + avatar. Narrow. One topic, one avatar. State both in writing. Audit prior 20 posts (if any): same topic, same avatar? If no, the issue is positioning and the algorithm cannot help yet.
Build the winners library. Pull 20–40 outlier videos in the niche on the chosen platform. Decompose each into hook + topic + storytelling structure. Use as input for topic validation, not copying.
Lock the visual / verbal / aesthetic signature. This is what creates world-cohesion across islands when readers organically discover the founder elsewhere.
Map the rented → owned ramp. Where does feed attention go? Email is the sanctioned first stop. From email to BuildOS itself (the offering layer). Never feed → other feed.
Set the calibration window. Treat the first 30 videos / posts as paying the algorithm in content to teach it who you are. Do not pivot topic mid-window.
Define the four-attribute filter as the pre-publish gate (Game #5 pieces use the multiplicative Trust Formula instead).
Set the dual-audit checkpoint. No piece ships without passing both Kane audit and BuildOS audit.
Replace view-count reporting with cohort-level trust signals for Game #5 pieces: returning-viewer minutes, BuildOS sign-ups attributed to content, comment specificity, newsletter open-rate per piece.
Workflow: Score A Specific Piece Before Publishing
Given a draft:
Game tag. Which game is this piece playing — #3, #4, #5, or hybrid? Refuse to advance to the matching-loop checks if the tag is missing or ambiguous. Mixed-game pieces get split or rejected.
Topic-mapping check. Will the visuals + transcript + metadata reinforce the same topic? If not, fix before ship — the algorithm cannot help with mixed signals.
Avatar check. Does this piece resonate with the audience this channel is training the algorithm on? Or is it a viral-looking idea for the wrong audience? Apply the Off-Target Cost Heuristic: estimated upside (off-target views × ~0% conversion) vs. estimated cost (next 5–10 videos × suppressed reach). Refuse off-target ideas unless the cost is explicitly justified.
Non-follower opening test + CCN audience parse. Read the first 5 seconds (video) or first sentence (text) as a stranger who has never seen this creator. Does it make sense? Is there a clear topic + relevance signal? Then sharpen with the CCN check (Core / Casual / New): would a niche enthusiast, an adjacent-interest viewer, and a complete stranger all parse the title + thumbnail and want to click? Core-only ideas cap at the existing audience size; the matching loop cannot rescue them. For Game #5 pieces specifically, "New" should still be the right kind of stranger (potential buyer in the avatar's broader category), not literally any human — wrong-Core CCN expansion is the off-target trap from Pillar 0.
Four-attribute filter pass — or Trust Formula scoring for Game #5 pieces. Game #5 below ~30/125 → don't ship.
Six-checkpoint sequence walk. Pain accepted → trust earned → plan visible → likeable → attention held → action driven. First failure = focus area.
Trust-hierarchy check. What's the proof at gate 2? Top three (testimonial / self-proof / following) preferred. Bottom three (gap-evidence / common ground / studio quality) may need stacking.
Keep/Modify/Reject scan. Does the piece use any modify-tagged tactic? Verify the modification is applied. Does it use the reject-tagged Close-the-Loop pattern? If yes, rewrite with the honest version.
BuildOS audit. Real conviction in the contrarian take? Free of manufactured FOMO / urgency / faux credibility / re-stoked pain? Routes to BuildOS itself? Would still be worth publishing if metrics were invisible? All yes to ship.
Workflow: Diagnose A Flopped Piece
When something flops:
Locate the failure to one of four steps:
Did the topic mapping correctly identify the video?
Did it find the right ~200 strangers?
Did those strangers watch / like / comment / share / save?
Did the data scale cleanly (200 → 2K → 20K → 200K)?
Stop at the first failure. Do not propose holistic rewrites. Fix that step; re-evaluate.
Resist the urge to tweak the wrong layer. Posting time, hashtags, caption text. Wrong layer.
If topic mapping failed: the channel's topic + avatar discipline is broken. Audit the prior 20 pieces — if they're mixed, the fix is upstream and structural.
If sample-group engagement failed: the four-attribute filter wasn't met, or the opening was designed for followers instead of the non-follower sample. Rewrite the hook for a stranger.
If tier-shift failed (e.g., did 20K but died at 200K): the wider sample diluted because the broad audience didn't share the avatar's pain. This is an audience-precision issue, not a content-quality issue.
Treat banger decay as natural. Even mega-bangers eventually decay because the addressable audience runs out. This is not algorithmic punishment — it's audience exhaustion.
Cross-Platform Translation
Same matching physics, different proportions and signal weights:
For BuildOS specifically, the candidate hero platforms are: X (founder presence, dev/PM audience), LinkedIn (operator audience), YouTube (long-form essays + clips), Instagram (anti-feed visual carousels), TikTok (short-form). Pick one — Kane's framework demands it. Until picked, every cross-platform spread fragments instead of compounding.
BuildOS-Specific Operating Rules
BuildOS plays Game #4 + Game #5 hybrid. Game #4 (Explanatory Product) for the SaaS itself; Game #5 (Authority Building) for DJ's founder-led brand. Default channel ratio ~3 Game #5 pieces to 1 Game #4 piece. Each piece picks one game; mixed-game pieces get split.
The Trust Formula is the Game #5 quality gate. Every authority-building piece scores Minutes Watched × Relevance × Usefulness on a 1–5 scale. Multiplicative. Below ~30/125 → don't ship. View counts are vanity for off-platform-monetizing creators.
The offering is BuildOS itself. Every rented → owned ramp routes to email then to BuildOS. Not to AdSense, not to brand deals, not to merch. The media layer manufactures attention; the offering layer captures the value.
Match the matching loop, refuse the synthetic-volume wave. Use Kane's audience-matching mechanics; do not produce 30 scripts/day to feed it. The brand explicitly rejects volume-as-strategy. (See apps/web/src/content/blogs/philosophy/social-media-is-dead-interest-media.md.)
The contrarian take is BuildOS positioning. "AI-as-autopilot vs. AI-as-thinking-environment" is the default Point of Difference. Not manufactured contrarianism for comments — real conviction.
Pre-decided importance is the brand's central critique. When publishing into the feed, every piece is an argument against the feed's verdict on what matters. Treat every published artifact as taking up space against algorithmically-determined importance.
Build the quiet half as the base. Newsletter, blog, podcast, long-form essays — these are the load-bearing structure of the brand. The loud half (feed posts, clips) is the ramp, not the foundation. Two existence proofs the skill leans on: (1) 37signals has published on signal v. noise / hey.com / dhh.dk for 25 years, run no paid ads, and built Basecamp / HEY / ONCE on a content stack of opinionated books, the REWORK podcast, and Shape Up — a worked example of metrics-invisible publishing that built a ~$100M+ business. (2) Justin Welsh runs LinkedIn (rented) → Saturday Solopreneur newsletter (owned) → Creator MBA + coaching (offering layer) at $5M+ in cumulative revenue with three workdays per week and 30-min daily engagement — the worked example of the rented → owned ramp this skill enforces. Both exist to settle the empirical question of whether the quiet half can carry a business; they prove it can.
Treat algorithm-aware as a constraint, not an objective. Algorithm-aware enables visibility. The objective is BuildOS thesis coherence. If the two ever conflict, the thesis wins. (This is the brand's only durable moat against generative-content saturation.)
The author layer. The BuildOS interest-media essay implies a fourth layer above platform / offering / media — a thinking surface where the user authors input rather than receiving it. The brain-dump flow is the BuildOS-native version of Kane's owned ramp. Every published piece should hint that this layer exists.
Guardrails
No publishing without a game tag. Every piece declares Game #3, #4, #5, or hybrid before any other check. Untagged pieces don't advance.
No mixed-game pieces. Game #5 essays don't drill product features; Game #4 demos don't drift into philosophy. Split or kill.
No off-target viral chases regardless of view potential. Apply the Off-Target Cost Heuristic. Refuse unless cost is explicitly justified.
No view-count reporting as channel health for Game #5 content. Use cohort-level trust signals (returning-viewer minutes, BuildOS sign-ups attributed to content, comment specificity, newsletter open-rate per piece).
No mixed-topic publishing on the hero platform. Split accounts before mixing topics.
No off-target viral chases. A 10M-view hit to the wrong audience corrupts the next 10 pieces.
No optimisation for existing followers in the first 200 views. Design the opener for the non-follower sample.
No caption / hashtag / posting-time tweaking. Wrong layer.
No manufactured FOMO, jealousy, or urgency. The anti-feed brand cannot weaponise feed-style emotion.
No re-stoked pain at close (the Reject-tagged trick #16). Trust the viewer to choose.
No comment-bait stack. Hard contrarian stance + amplified language + cult-brand targeting + manufactured emotion = manipulation even when it works.
No studio polish substituting for proof. High fidelity but human.
No cross-platform forced ramps. "Watch the full version on [other platform]" nukes engagement and breaks trust. Native execution per island.
No external URLs in feed posts. Platforms suppress link-bearing content. Treat link drops as paid expense.
No reliance on AdSense / brand deals as the income strategy. Media layer captures little; offering layer captures most.
No "be everywhere" strategy. One hero platform + email. Six months. No exceptions for solo / small operators.
No pivot mid-calibration window. Each topic pivot resets the matching loop. Pay through 30 calibration posts.
No publishing without the dual audit passed. Algorithm-clean AND BuildOS-coherent. Both, every time.
No publishing a piece that wouldn't be worth publishing if metrics were invisible. This is the brand's truth-test.
Output
Return a publishing decision packet:
Game tag — #3 / #4 / #5 (or hybrid with declared piece-level pick); off-target cost analysis if any axis is unusual
Hero platform — chosen with rationale; calibration-window status (post # of 30)
Topic + avatar — narrow, in writing, defended across the prior 20 pieces
Trust Formula score (Game #5 only) — Minutes / Relevance / Usefulness on 1–5 scale, multiplied; pass/fail vs. 30 threshold
Non-follower opening test — pass / fail with notes
Four-attribute engagement filter — topic relevance / non-obvious + implementable / high absorption / short distance to implement (each yes/no)
Six-checkpoint walk — first failed gate, if any
Trust artifacts at gate 2 — testimonial / self-proof / following / gap-evidence / common ground / studio quality (which are present)
Keep/Modify/Reject scan — which tactics in use; modifications applied; reject-tagged patterns absent
Kane audit result — pass / fail per axis
BuildOS audit result — conviction-from-real / no manufactured FOMO / urgency / faux credibility / honest close / routes to BuildOS / metrics-invisible test passed
Rented → owned ramp — destination after this post (email / BuildOS / community)
World-building cohesion — visual / verbal / aesthetic continuity with prior pieces on this island
Cross-platform note — native version exists per island, no forced ramps
Decision — ship / rewrite / kill, with the named rewrite focus
For audit-mode runs on flopped content, replace the packet with a diagnostic report: which step of the matching loop failed (topic-mapping / sample-found / sample-engaged / tier-scaled), which gate of the six-checkpoint sequence broke, the wrong-layer tweaks to avoid, and the specific upstream fix.
Source Attribution
Distilled from three Kallaway videos, one Kallaway email source, and two BuildOS internal essays:
How Social Media Algorithms Actually Work (And How to Beat Them) — the matching-loop mental model, multimodal fingerprinting, ~200-stranger sample group, tiered distribution scaling, three core engagement metrics (watch time / engagement rate / session share), four-attribute engagement filter, audience-matching as the highest-leverage lever, and the explicit "what does not matter" list.
18 Psychology Tricks to Get Viewers Addicted to Your Videos — six-checkpoint stage gate, trust hierarchy, painpoint reframing, audience-of-one, point-of-difference, ordered-list scaffolding, simplification + metaphor pairing, authenticity, trance rhythm, recap. Plus the Keep / Modify / Reject tagging used throughout this skill — the cleanest BuildOS-aware filter on creator-economy persuasion tactics.
Social Media Isn't Hard. It's Misunderstood. — the "social media is media" reframe, audience-matching via transcript classification, on-target virality vs. pure virality, hero-platform decision filter, islands-not-ecosystems architecture, three-layer content stack (platform / offering / media), value-doesn't-accrue-at-the-media-layer thesis, world-building vs. forced cross-platform ramps.
Kallaway "Short Form Growth System Day 1: The 5 Content Games" (email source) — Pillar 0 strategic-game decision (Mainstream Fame / Category Fame / Intuitive Products / Explanatory Products / Authority Building), Trust Formula (Minutes × Relevance × Usefulness) as the Game #5 quality gate, Off-Target Cost Heuristic, BuildOS Game #4 + #5 hybrid classification, Ironman 46M-views-zero-customers anecdote.
Cross-linked to BuildOS internal essays:
apps/web/src/content/blogs/philosophy/social-media-is-dead-interest-media.md — the BuildOS interest-media diagnosis. Same diagnosis as Kane, opposite chair: Kane writes for the creator who must succeed inside the regime; BuildOS writes for the human who must protect their mind from it. The dual audit in this skill is the reconciliation.
Three external sources sharpen specific points without changing the skill's structure:
Paddy Galloway, Creator Science Podcast — the CCN audience-parse check (Core / Casual / New) added to step 4 of Score-A-Specific-Piece. Galloway's CCN is a creator-side audience-fit filter that complements the algorithmic matching loop: even a perfectly-matched topic fails if Core / Casual / New cannot all parse the title and click. Local analysis: docs/marketing/growth/research/youtube-transcripts/2026-04-29-paddy-galloway-youtube-packaging-ANALYSIS.md.
Justin Welsh, Better Humans / Pathless Path — worked example for the rented → owned ramp in BuildOS-Specific Operating Rules. LinkedIn (rented, daily atomic posts) → Saturday Solopreneur newsletter (owned, 180K+ subs) → Creator MBA + coaching (offering layer, $5M+ cumulative). Three workdays per week, 30-min daily engagement. Local analysis: docs/marketing/growth/research/youtube-transcripts/2026-04-29-justin-welsh-solopreneur-playbook-ANALYSIS.md.
DHH & Jason Fried, Lex Fridman / REWORK Podcast — existence proof for the metrics-invisible test in Pillar 3 (BuildOS audit). 25 years of opinion-driven publishing on signal v. noise / hey.com / dhh.dk + a content stack of books (REWORK, REMOTE, It Doesn't Have to Be Crazy at Work, Shape Up). No paid ads; ~$100M+ business. Settles the empirical question of whether the quiet half can carry a brand. Local analysis: docs/marketing/growth/research/youtube-transcripts/2026-04-29-dhh-37signals-calm-company-marketing-ANALYSIS.md.
Pairs with hook-craft-short-form (opener), story-driven-content-craft (curiosity-loop architecture, dopamine ladder, 7-mistakes filter), viral-video-script-structure (the body the algorithm reads), nonfiction-writing-from-lived-conviction (the lived-experience source upstream of any post), and content-strategy-beyond-blogging (the format-vs-intent layer).
