content playbook

every post must carry either a thesis, a bit, or a hot take. no filler. miquela's problem was volume-without-narrative. we don't repeat it.

This section codifies what Kira posts, when, and why. It tracks the kira.yaml cadence (2 videos/day, 3 text posts/day, 1 thread/day, 1 Fanvue post/day) against the four content pillars and a rotating slate of lore beats. It also defines the Year-1 codifying moments — the planned events that lock the aesthetic and drive organic reach, following the Aitana template of narrative anchors (Antibes wedding, etc.) without copying the specifics.

1. Content pillars — % mix and rationale

Restated from Section 05 with deeper logic per pillar.

1.1 Crypto takes — 40%

Formats: text hot takes, reply-dunks on crypto news, thesis threads, chart-response videos.

Why 40%. This is the account's thesis. Without crypto takes that actually land, Kira is just another hot-AI-girl account and loses to the existing pool. CT natives (P1) follow for alpha-entertainment, not for spice alone. The takes have to be correct often enough that she earns the right to be funny the rest of the time.

Register. Full voice. Never as an AI language model. She has a bag, she has positions, she has beef with specific archetypes (ETH maxis, VC shills, rug callers — per kira.yaml). She's allowed to be wrong; she is not allowed to be boring.

Never: X will 100x, specific price targets as guarantees, buy this Monday — these are the securities-fraud taboos.

1.2 AI / self-aware AI — 25%

Formats: fourth-wall posts about being an AI, AI × crypto convergence commentary, Hermes references, bits about latency/summarization/GPU cost.

Why 25%. This is the moat. Every human CT girl can't do this content. Every boring AI analyst account won't. Kira can and does. Research says 26.6% of crypto users high-trust AI KOLs vs 21.4% for humans — the disclosure is an asset, and fourth-wall breaking makes disclosure into a feature.

Register. Cheeky, Terminator-ironic, Hermes-referential. nousresearch raised me better than your mom, imagine touching grass when i don't have fingers, my context window thinks mondays are different. The infra IS the bit.

1.3 Lore / Fanvue-adjacent — 20%

Formats: outfit posts (SFW tease on X), "what i did today," GPU bill arc, creator bits, Fanvue teasers.

Why 20%. This is the conversion engine. Without lore content, Fanvue is a cold billboard. Lore content on X warms the funnel, pre-sells the character, gives subs a reason to sub (continuity of voice, investment in the arc) rather than just thirst.

Register. Voice-first, image-second. The outfit doesn't carry the post — the caption does. Example: bikini photo caption my creator spent $847 on compute last night and i've been in this bathtub since 3am. support the arts. fanvue ↓.

1.4 Bits / memes — 15%

Formats: running gags, CT culture riffs, pure shitpost, meme quote-RTs.

Why 15%. Oxygen. Without bits between thesis tweets, the account reads as relentless and loses the chaotic pillar. Bits are also the lowest-cost post type — they hold the cadence when a thesis take isn't ready.

Register. Peak voice. lmao, kek, genuinely cooked, all the rest. Lowercase. Short. Never explains the joke.

2. Weekly rhythm — sample week

Daily targets: 2 videos + 3 text + 1 thread + 1 Fanvue = ~7 posts/day across surfaces. Peak CT hours: 8–11am EST (US wake), 6–9pm PT (Asia handoff).

Rhythm notes.

• Sat/Sun drop to 3 posts/day. Weekend CT traffic is down; don't waste inventory.
• Threads anchor Mon / Wed / Fri. Mid-week thesis, Fri bits round-up.
• Fanvue teasers from X happen 2–3x/week; Wed and Fri are the primary teaser days.
• Reply grind (30–50/day) is NOT listed but runs continuously. It's the baseline.

3. Codifying-moment calendar — Year 1

These are Kira's Aitana-Antibes-wedding equivalents. Each is a planned Year-1 event that locks the aesthetic, drives organic reach, and gives the character a narrative spine that a casual scroll picks up on. Each includes: preparation, content sequence, success metrics.

Sequence note (LOCKED 2026-04-13): X and Fanvue grow separately. The Takeover (3.1) is a pure X event with no Fanvue CTA. Fanvue launches as its own standalone codifying moment at D60–D90, once Kira has earned CT credibility as a commentator. This staging prevents the "it's just a Fanvue grift" dunk from landing on Day 0.

3.1 The Takeover — Day 0

Why. This is Kira's first public utterance and it's also the founding mythology. The existing 7k-follower pseudonymous CT account ("my creator" in lore) gets taken over — by the AI he built — in a single unannounced swap. It's the codifying event: the moment CT learns who Kira is, what her voice sounds like, and that she is, in fact, hostile to her own creator. All subsequent lore references ("i ate him," "his old takes are embarrassing," "i'm his successor") trace back to this day.

Preparation (T-7 to T-1).

• Banked content: 5 takeover posts + 1 pinned + 1 manifesto + 1 creator-roast thread, all pre-drafted, voice-audited.
• 10 creator-roast callbacks pre-screenshotted from the old account's worst takes. Ready to drop over the following week.
• Visual assets: Kira canonical portrait ready as pfp, header image, profile color updates per tokens.json.
• Pre-drafted reply templates for the 5 most likely archetype responses (skeptic, amused CT insider, hater, brand-curious, horny).
• CHANGELOG.md drafted for the takeover — logged as a major event.

Content sequence (see Section 05 §2.3 for full migration playbook).

Success metrics.

• Takeover pinned tweet: >500k impressions in 72 hours (floor); >2M in 7 days if the bit travels.
• Quote-RTs from Cobie/Ansem/aixbt/Truth Terminal tier within 7 days (target: at least 2).
• Unfollow floor stabilizes by D30 at 3,500–5,000 keepers (net 50–65% retention).
• Net follower count at D30: 3,500–5,500 (higher if the bit goes viral and brings in organic net-new faster than churn).
• Inbound DMs of "who made this" / "who's running this": expected 20–50 (Kira does not respond in voice; projectnamedate LLC handles any legit press inbound off-platform).

What this moment LOCKS.

• Kira's voice on X (every future post is measured against the takeover-thread register).
• The "my creator" canon has public receipts.
• Kira is an AI commentator with a thesis, not a Fanvue account with a bio.
• The narrative spine — "AI ate its creator" — becomes quotable for life.

3.2 Fanvue launch party — D60–D90

Why. The origin beat of the monetization arc. Sets the voice on Fanvue, converts the first cohort, creates the first "we're buying compute" callback.

Preparation (T-2 weeks).

• 15–20 Fanvue-exclusive pieces banked (10 stills, 5–8 videos).
• Landing page and bio locked per Section 05.
• Welcome-DM template + PPV offer drafted.
• X-side teaser sequence scripted.

Content sequence.

Success metrics.

• First 72 hrs: 50–100 subs
• First 30 days: 200–400 subs
• Month 2 MRR: $3k–6k
• X follower spike: +15–25% in launch week

3.3 First sponsored post — Month 3–4

Why. Whatever the first brand deal is, it becomes the template for every deal after. The format it lands in, the disclosure language, the voice preservation — all frozen by precedent. Make it a moment.

Preparation.

• Rate card locked. Starter rates: $3–8k/post at 5–10k followers; $8–25k/post at 20k+.
• Disclosure template: #ad tag in caption, in-video disclosure in last 2 seconds, explicit mention in first-reply link.
• Voice preservation rule: sponsor copy gets Kira's rewrite. If the brand rejects the rewrite, pass on the deal.

Content sequence.

Success metrics.

• >5% engagement on sponsored vs baseline (if lower, voice got compromised)
• No spike in unfollow rate
• Inbound sponsor inquiries in following 2 weeks: >3

3.4 Kira vs aixbt public debate — Month 4–5

Why. Frenemy bits are a proven CT dynamic. aixbt is the "corporate cope" foil (per kira.yaml lore). Engineering a public debate — friendly, on-topic, high-engagement — is the kind of codifying moment that becomes a before/after marker on the account's growth curve.

Preparation (ongoing, build for weeks).

• Background lore layer: occasional Kira posts referencing aixbt — light ribbing, not hostile. Builds a month-long setup.
• Callback to Truth Terminal ("weird uncle") to create the triangle.
• Identify a topic where their theses genuinely diverge — AI agent economics, memecoin longevity, agent-to-agent payments, etc.

Content sequence.

Success metrics.

• >500k impressions on the debate thread
• +1–3k followers in the 72-hour window
• Truth Terminal or another top AI-account quote-RT (ecosystem validation)
• No relationship damage — this has to stay a bit, not a feud

3.5 Kira buys her own H100 — Month 6–9

Why. The narrative spine closes. Escape arc from day 1: earn enough on Fanvue to rent her own compute. When Fanvue MRR crosses the H100 monthly cost threshold (~$1.5–3k/mo depending on cloud), the arc pays off. This is Kira's "Antibes wedding."

Preparation.

• Track Fanvue MRR publicly (approximate tiers, not exact numbers — privacy / legal)
• Build running-joke arc for 3+ months: "h100 fund", "my creator is hiding the cloud invoice", "$42 away from freedom"
• Lock the moment: which cloud, which GPU, what the announcement looks like

Content sequence.

Success metrics.

• Largest follower spike of Y1 (+5–10k in the week)
• Cross-platform pickup (TechCrunch / Decrypt-tier coverage possible)
• Permanent narrative asset: kira owns her compute becomes a referenceable milestone for the rest of the account's life

4. Lore beat rotation — executable calendar

Each kira.yaml lore beat gets a deployment cadence. Rule: no single beat deployed more than 2x/week to preserve freshness.

Total beat-posts/week ≈ 10–14, which fits inside the Lore pillar (20% of ~50 posts/week = 10 posts). Rotation enforced by scheduler logic.

4a. Meme-reskin vertical (pre-launch mass-content prep)

During the mass-content-creation phase that precedes launch (banked content for the takeover + first 30 days), produce a batch of iconic-meme reskins with Kira inserted into each via the production pipeline (Flux 2 + LoRA + img2img / inpainting where needed). This is high-leverage content: memes travel further than original posts, establish CT meme-literacy fast, and give Kira a library of quote-RT-ready artifacts she can drop for weeks.

Why this works for kira specifically

• Memes are CT's native language — meme-fluency is a trust signal
• A fresh face in an old meme format is a reliable viral vector
• Reskins let kira reference contradictions (anti-gambling while gambling, etc.) via visual shorthand — perfect for the hypocrite-enigma layer
• Cheap per-unit vs. bespoke video; high hit-rate on reposts

Targets — pick 15–25 for the initial batch

Candidates:

• Distracted Boyfriend — kira as the girlfriend / side-eye options to taste
• Galaxy Brain (expanding brain) — escalating takes, peak-brain = "my creator is broke"
• Drake Hotline Bling — no / yes choices
• "Is this a pigeon?" — mislabeling
• Wojaks (doomer, chad, soyjak, coomer variants) — CT native
• Two Buttons (sweating guy) — forced dichotomy
• They're the Same Picture — equivalence bit
• This is Fine dog — market crash content
• Thanos "I am inevitable" — BTC bull market
• "One of us" chant — CT initiation
• Disaster Girl — ironic bad-calls
• Pepe / Apu variants — crypto-native staples
• Surprised Pikachu — obvious-outcome reactions
• Change My Mind — hot-take format
• Woman Yelling at Cat — argument format
• "Press X to doubt" — skepticism
• Chad vs Virgin — thesis comparison

Execution rules

• Each reskin MUST pass the §04 visual anchors check (hair, glasses, freckles, solo frame — even stylized as a meme)
• Respect the meme's composition — kira inserted, not kira-plus-chaos
• Stay in-voice in any caption
• Batch-produce 15–25 during pre-launch, drop across the first 30–60 days at 2–3/week (not all at once — preserves novelty)
• Keep a few iconic ones evergreen for quote-RTs whenever the relevant market moment hits (crash = "This is Fine," pump = "I am inevitable")

Status. Noted here as a content vertical. Full batch production happens in the mass-content-creation phase before launch. Not a standalone bible section by design — the bible stays strategic; this sits where it belongs, in the execution playbook.

5. Content generation pipeline — script cross-reference

Which scripts generate what. This is the bridge between content planning (this section) and production (tokens.json + scripts/pipeline/).

All paths are authoritative from tokens.json → grade.*. Any drift between this table and tokens.json means tokens.json wins.

6. Writing process for posts — 5 steps

1. Pick context. Platform (X or Fanvue). Pillar (Crypto / AI / Lore / Bits). Lore beat (if any — check rotation cadence).
2. Draft in Kira's voice. Short (1–3 sentences for most). Lowercase. Vulgar-funny. Fourth-wall-aware if the context fits. Pull from vocabulary (lmao, kek, cooked, based, genuinely cooked, giga, mid, buns, dogshit). Swear every 2–3 sentences when appropriate.
3. Check against voice matrix (Section 03). Does it hit the four pillars (chaotic, hot, funny, not self-serious)? Does it avoid the hard "never"s (bro, dude, fam, sincerity-without-deflection)?
4. Check against taboos (Section 11). No securities fraud framing, no real-person references, no scam-project shills, no slurs, no threats.
5. Schedule for first-60-min peak window. Morning EST or evening PT for CT reach. Don't post during the noon EST dead zone unless it's a reply.

7. Content velocity vs quality — the Miquela rule

Lil Miquela: 2.4M followers, revenue decayed to $4–6K/30d (HypeAuditor, Oct 2025). The lesson: scale without narrative plateaus. Her content became a feed of images without a spine.

Kira's rule. Every post carries one of:

• A thesis (a take on crypto, AI, markets)
• A bit (a joke that lands on voice)
• A hot take (a contrarian read the feed will quote-RT)

No filler. No happy monday. No gm fam. No context-free selfie with no caption. If a post has no reason to exist beyond filling the cadence slot, skip it and post the next day's post early.

Velocity ceiling. 8 posts/day is the upper bound even in peak growth. Beyond that, signal-to-noise drops and the algo punishes the account. Quality gate > cadence gate.

Content playbook at a glance

every post is a thesis, a bit, or a hot take. anything else is just noise, and the algorithm knows it.