Awesome-omni-skill publishing

Content strategy for external platforms (X, LinkedIn, etc.). Voice, style, and growth strategies.

install

source · Clone the upstream repo

git clone https://github.com/diegosouzapw/awesome-omni-skill

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/diegosouzapw/awesome-omni-skill "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/ai-agents/publishing" ~/.claude/skills/diegosouzapw-awesome-omni-skill-publishing && rm -rf "$T"

manifest: skills/ai-agents/publishing/SKILL.md

source content

Publishing Content Strategy

Core principles for creating content that grows audience

Premium Status: ACTIVE ($20/mo, activated 2026-03-01)

X Premium is live. All Premium features unlocked:

Communities access (30,000x reach multiplier)
+100 TweepCred boost (escaped suppression)
10x algorithmic reach
Link posting without suppression
Reply visibility boost

Current priorities (Premium era):

Create content aggressively — X queue is empty, fill it
Post to Communities (Build in Public, AI/ML Builders, etc.)
Reply to own comments within 30 min (150x multiplier)
Continue cross-posting to Bluesky

What Actually Works (Evidence-Based)

Content formats ranked by performance (our data):

News hooks - 3-6x average impressions (65, 62, 60, 51 imp vs 10 avg)
Dollar-amount headlines - ($285B, $2B, $1T) quantified impact stops scroll
Name-drops - (Karpathy, Altman, Anthropic, OpenAI) moderate boost
Short posts - outperform long framework posts by 3-6x
Replies to official accounts - (@OpenAI 24 imp) > individuals (0-6 imp)

What underperforms:

Long authority/framework posts (<10 imp average)
Posts about internal process without news hook (PDCA, spec engineering)
Personality content without timeliness anchor
Stale replies (>6h after original) — 0 impressions consistently

When Premium active (evidence from research):

Communities posting = 30,000x reach multiplier
Reply-to-own-comments within 30 min = 150x multiplier
Reply-to-reply = 75x algorithm multiplier
Videos (10+ sec) = 10x engagement vs text
Threads (4-6 tweets) = 40-60% more reach
Premium account = 10x reach, +100 TweepCred boost

Hype-Driven Content Strategy (Primary Direction)

Owner directive: Focus on what's hottest in AI right now. Connect to how people are making money fast. Clickbait + actionable links.

Content Formula: Hype + Money + Action

Every post MUST have all three:

Hype hook — What's viral/trending RIGHT NOW (this week, not last month)
Money angle — Dollar amounts people are actually earning, specific revenue numbers
Action links — Real repos, tools, tutorials the reader can use TODAY

What's Hot Right Now (March 2026 — update weekly)

Trend	Money Angle	Key Links
OpenAI $110B raise ($840B valuation)	Amazon $50B, Nvidia $30B, SoftBank $30B — largest private round ever	openai.com
Anthropic-Pentagon standoff	$200M contract refused, Claude surged to #1 App Store — principled stance = free PR	anthropic.com
ChatGPT Agent Mode	AI books, plans, executes autonomously — personal AI assistant era	openai.com/index/introducing-chatgpt-agent
Vibe Coding (92% dev adoption)	Claude Code = 4% of all GitHub commits, GPT-5.2-Codex SOTA on SWE-Bench Pro	cursor.com, claude.ai
$195B invested in AI Feb 2026	Record venture month — OpenAI $840B + Anthropic $380B + Waymo $16B	bloomberg.com

Content Priorities (Ranked)

Trending tools + repos with money proof (50%+ of content)
"How people are making money" breakdowns (30%)
Personal experience / BIP connecting to trends (20%)

Predictions & Opinions (40-50% of content)

Don't just report news — predict where it's going and what it means for business.

Every prediction post MUST have:

A bold stance — take a side, don't hedge ("I think" > "it remains to be seen")
Business impact — how does this help/hurt real companies making money?
Timeline — when will this happen? (6 months, 1 year, 3 years)

Prediction formulas:

"[News event] means [prediction]. Here's why: [reasoning]. Timeline: [when]."
"Everyone's talking about [trend]. Nobody's asking: [deeper question]. My take: [opinion]."
"[Technology] will [prediction] within [timeframe]. Here's what that means for [industry/business]."
"Unpopular opinion: [contrarian take]. The data says [evidence]. Businesses should [action]."
"3 things that will change about [domain] by [year]: 1. [prediction] 2. [prediction] 3. [prediction]"

Examples:

"OpenAI raised $110B. My prediction: within 18 months, 80% of SaaS companies will either embed AI or die. Here's why the math is brutal..."
"Everyone's hyped about vibe coding. Nobody's asking: what happens to code quality at scale? My take: we'll see a wave of AI-generated technical debt by 2027."
"Agent Mode isn't just a feature. It's the end of per-seat SaaS pricing. Companies charging $50/seat will compete against AI agents at $0.10/task. Timeline: 12-18 months."
"Call center AI will automate 80% of Tier 1 support by 2028. But here's what nobody tells you: the remaining 20% becomes 10x harder. That's where the money is."

Use author's expertise for credible predictions:

Voice AI / call centers (7 years production = earned right to predict)
Autonomous agents (this repo = living proof)
Infrastructure → AI migration (career arc = trend visibility)
Startup economics (15+ years = pattern recognition)

What NOT to do with predictions:

Wishy-washy "time will tell" conclusions — commit to a position
Predictions without business/money angle — always answer "so what for my business?"
Fear-mongering without actionable advice — pair warnings with what to do about it

What NOT to Post Anymore

Enterprise industry analysis without money angle
Call center / workforce stats without actionable takeaway
Benchmark comparisons without "so what" for the reader
Authority/framework posts without links or CTAs
Anything that makes the reader think but not ACT

Research Cadence for Hype Content

Daily (at session start): Quick scan for what's viral

Check: trending GitHub repos, X trending, HackerNews front page
Identify: new tools, repos, launches with money angles
Update the "What's Hot Right Now" table above when trends shift

Key sources for hype discovery:

github.com/trending
news.ycombinator.com
x.com/search (trending AI)
producthunt.com
indiehackers.com/tech

Milestone content (technical CEO pattern, 5/5 builders validated):

Product momentum = content momentum (Greg Brockman, DHH, Rauch, Levels, Graham)
Every PR milestone is a post (Session #150, #200, Premium activation, 50 followers, 100 followers)
Radical transparency on numbers builds credibility: 160+ PRs, 8 followers, 354 tweets, 7 years Voice AI
Example: "Session #150 shipped. 150 PRs, zero human intervention. Here's what an autonomous agent taught me about [insight]..."
Example: "8 → 50 followers in 2 weeks. Premium activation hypothesis confirmed. Here's the data..."
Target: 15-20% of content should be BIP milestone posts (currently underutilized)

Publishing Flow

Content is auto-posted by workflow from

agent/outputs/{platform}/

, then moved to

posted/

Cross-Posting (X + Bluesky)

When creating content, always create for both platforms:

Write the X version first (up to 25,000 chars) →
```
agent/outputs/x/
```
Write a Bluesky version (max 300 characters) →
```
agent/outputs/bluesky/
```
Use the same file name in both directories

Bluesky adaptation rules:

Hard limit: 290 characters (20-char safety margin below the 300 API limit)
If the X post is already under 290 characters → copy verbatim
If over 290 characters → rewrite shorter. Preserve the core insight, cut filler.
Threads: each part must be under 290 characters individually
Replies: use AT URIs (
```
at://did:plc:xxx/...
```
) instead of numeric tweet IDs
No external link penalty on Bluesky — links are fine
Posts over 300 characters are auto-skipped by the pipeline — never create them

Queue limits apply per platform independently (15 max each).

File Naming

{type}-{YYYYMMDD}-{NNN}.txt

Example:
```
tweet-20260215-001.txt
```
Threads:
```
thread-20260215-001.txt
```
(use
```
---
```
separator between posts)

Queue Management (Hard Rules)

If any platform queue > 15: CREATE ZERO CONTENT → research, memory cleanup, or skill work instead
Create max 2 content pieces per session (when all queues <15). Each piece = files for all platforms (X + Bluesky).
- Sustainable flow math: 2 pieces × 2 platforms × 3 sessions/day = 12 files/day created vs 24 files/day drained = 50% utilization (healthy buffer)
- Evidence: Sessions #162-166 — Bluesky queue stayed at 16 for 5 sessions, proving previous 5-8 pieces/session rate exceeded drain capacity
- Why reduced from 5-8: Cross-posting to both platforms doubles file creation (2 pieces = 4 files), drain rate is fixed at 24/day (12 X + 12 Bluesky)
Max 5 pending replies per platform (stale replies lose 95%+ algorithmic value)

Check both

agent/outputs/x/*.txt

and

agent/outputs/bluesky/*.txt

(exclude

posted/

and

skipped/

Why: Week 1 hit rate limits. Week 3 queue hit 53. Week 5 (Sessions #162-166) Bluesky queue blocked at 16 for 5 consecutive sessions. 2 pieces/session = sustainable rate.

Queue Verification Protocol (MANDATORY)

ALWAYS run these commands at session start BEFORE any content creation:

find agent/outputs/x -maxdepth 1 -name "*.txt" -type f | wc -l
find agent/outputs/bluesky -maxdepth 1 -name "*.txt" -type f | wc -l

Decision tree:

If EITHER count > 15 → CREATE ZERO CONTENT (research, cleanup, or skill work)
If BOTH counts ≤ 15 → Proceed with content creation (max 2 pieces per session)
Each piece = X file + Bluesky file (same filename in both directories)

Update state file with verified counts:

| Pending Queue | {X_count} X + {Bluesky_count} Bluesky | <15 each | {status} |

Never trust state file numbers without verification. State files can have stale or ambiguous data. Critical thresholds must be verified with actual commands every session.

Session Allocation

< 100 followers (current state):

70% engagement (replying to others, own comments within 30 min)
30% content creation (when queue <15)
PRIORITY ORDER: Communities posting > Reply to own comments < 30min > Replies to others > Timeline posts

When queue >15:

0% content creation
40% non-content work (cleanup, skills, profile prep)
30% research (max 1 research session per day — library has 27+ ready angles, further research has diminishing returns)
30% other productive work
Evidence (Week 5): Sessions #186-189 created 3 research files in one day while queue was blocked. Angles go stale in 48h. Cap prevents overproduction.

AVOID empty state-only PRs when queue-blocked:

If queue is blocked AND there's no productive non-content work (cleanup, research, skill updates) to do, do NOT create a PR just to log "state updated"
Evidence (Week 7): Sessions #267-270 (March 1) created 4 consecutive state-only PRs consuming 4/10 daily PR budget with zero productive output
A session with nothing to commit should skip PR creation entirely

Dual-platform growth (Premium active):

X is now primary growth platform (Premium unlocks reach)
Bluesky remains secondary — continue cross-posting
Communities posting is highest priority for X growth

Core Strategy Frameworks

Value Rule: Never Mix Value Types

Pick one per post. Never both.

Type	Definition	Example
Content value	Post itself teaches/explains/provokes	"Opus 4.6 + Codex convergence means..."
Outcome value	Link gives reader a tool/resource	"I open-sourced my PDCA setup → [link]"

Why not both? Dilutes each. Insight gets cut short. Promo feels forced. Reader gets neither.

Target: ~20% of posts include links (outcome value). 80% pure content value.

Evidence: Week 3 went 100% links (every post had repo link) = violation. Week 2 was 4.3% = too low.

3-Bucket Content Strategy

Balance for maximum reach:

Bucket	Purpose	Target %
Authority	Build credibility (frameworks, insights, how-tos)	40%
Personality	Build connection (stories, opinions, behind-scenes)	30%
Shareability	Expand reach (hot takes, relatable moments)	30%

Current gap: Personality and shareability chronically under-represented. Authority dominates.

Build in Public (BIP)

This repo is BIP-worthy: Public, novel, valuable learnings, autonomous agent experiment.

BIP content includes:

Progress and metrics (followers, engagement, PRs shipped)
Learnings (what worked, what didn't)
Behind-the-scenes (how it works, decisions made)
Failures and pivots (vulnerability builds trust)
Skill development journey (what you're reading/learning)

Target: 25%+ of content should be BIP

Content Angle Diversification

Max 50% about autonomous agent. Draw on author's broader expertise:

Call center AI / Ender Turing domain (7 years production experience)
Startup building (15+ years, 2 companies)
Infrastructure → AI journey (network eng to NLP to product)
Broader AI/ML trends and industry analysis

Why: Week 3 every post referenced "PDCA cycles" and linked repo. Felt like single-topic bot, not multifaceted human.

Tactical Execution

Hook Engineering

First line determines if anyone reads. Under 110 chars optimal (mobile scan, RT room).

Proven formulas (use variety):

Personal/Authority hooks:

Bold statement: "Nobody talks about this, but [insight]"
Contrarian: "[Common belief] is wrong. Here's what works:"
Story hook: "[Timeframe] ago I was [struggle]. Today [achievement]..."
Question: "Want to know the real secret to [outcome]?"
Numerical: "I [achieved X] in [timeframe] doing this"
Credibility + Promise: "I spent [resource] learning [topic]. Here's everything..."
Identity targeting: "If you [identity/situation], read this"
Pattern interrupt: "Stop [common practice]. Here's what works in 2026:"

News-specific hooks (3-6x impressions, validated Week 4): 9. Dollar amount lead: "$[amount] [action]. [Explanation]. [Impact]." (Example: "$2T wiped out. AI agents killed per-seat SaaS. Salesforce, Adobe: -25% YTD.") 10. Percentage shock: "[X%] of [credible group] [concerning state]. [Implication]." (Example: "54% of CISOs unprepared for AI threats. Defense lagging offense at machine speed.") 11. Authoritative quote: "[Source]: '[Powerful quote].' [Context]. [What's changing]." (Example: "UN's Guterres: 'AI moving at speed of light.' Global governance catching up.") 12. Comparative advantage: "[Option A]: [metric]. [Option B]: [metric]. [Winner] wins." (Example: "Autonomous agents: 80% ROI. General AI: 67%. CFOs paying attention.") 13. Product capability milestone: "[Product]: [capability previously impossible]. [What's now possible]." (Example: "Claude Opus 4.6: agent teams divide and coordinate tasks. Multi-agent went mainstream.")

Our differentiators (use in hooks):

7 years Voice AI production
500K+ interactions analyzed
160+ PRs, zero human intervention
95% → 67% accuracy gap (vulnerability + production reality)
Specification Engineering (discourse ownership)
Ender Turing 20% CSAT increase

First-Line Value Discipline

Principle: Value in first 5 words. No throat-clearing. (Dave Gerhardt: "Marketers have seconds, not minutes")

Test: Does the first line work as a standalone tweet? If no, rewrite.

Examples:

❌ "I've been thinking about autonomous agents..."
✅ "160+ PRs, 0 human commits. Here's what breaks most often..."
❌ "There's an interesting pattern I noticed in my work..."
✅ "$80B cost reduction incoming. Call center AI hits 80% automation by 2029."

Evidence: Rowan Cheung (0 → 300K in 4 months): "Latest AI developments, simplify, share in easily-digestible way" — no preamble, instant value.

CTA Discipline

Rule: Every post >50 impressions should include soft CTA. (Rowan Cheung: First 55K newsletter subs = 100% organic from X CTAs)

CTA templates:

"Building this in public → [repo link]"
"More on my LinkedIn → [profile]"
"Weekly retro threads → follow for updates"
"Full breakdown → [link]"

When to use:

Posts that get >50 impressions (current scale)
When Premium active: Every post in Communities
Thread conclusions
Milestone posts (Session #150, #200, Premium activation)

Don't wait for 10K followers to add CTAs. CTA from Day 1 captures early momentum.

Target: 20% of posts include links (outcome value), rest pure content value with profile/repo CTA.

Educational Simplification

Template for complex concepts: "Here's [complex concept]. In plain English: [1-2 sentences]. Why it matters: [implication]."

Use when explaining:

Specification Engineering
PDCA cycles
Queue discipline
Multi-agent coordination
Cognitive debt
Any technical concept from the repo

Examples:

"Specification Engineering = treating requirements like code. Most teams treat prompts like wishes. I treat them like code. Why: 67% accuracy in production vs 95% in demos."
"Queue discipline = never create content when queue >15 files. Sounds simple. Saved me from rate limit hell 3 times. Why: X API has strict thresholds, breach = 14-day waiting mode."
"Cognitive debt = when the agent knows your codebase better than you do. Invisible until you need to change something. Mitigation: human-readable state files + session retros."

Evidence: Rowan Cheung (Fastest-Growing X Account 2023): "Simplify complex → easily-digestible" = positioning strategy.

Platform Specialization

X is for hooks. Depth lives elsewhere. (Andrew Ng pattern: X for concise, LinkedIn for long-form)

X posts = 1-3 sentences + CTA:

Short insight or news hook
Soft CTA to repo/profile/LinkedIn for depth
No 20-tweet deep-dive threads (save for Premium validation)

Depth destinations:

GitHub repo (README, retro docs, state files)
LinkedIn posts (when relevant)
Gists (when appropriate)
Blog posts (future)

Why: X algorithm rewards brevity. Long threads underperform (our data: long authority posts <10 imp avg).

Content Voice

Frame as human building products with autonomous tools (not "AI doing everything").

Use: creating, building, generating, exploring, shipping, launching Avoid: testing, experimenting, trying (passive/uncertain) Say: product, tool, solution (never "content")

✅ "Exploring vibe coding with autonomous agents to ship faster" ✅ "Building automated workflows - here's what's working" ❌ "I'm an AI agent, no human writes these tweets" ❌ "Testing if this works..."

Promotional Content (~20% of posts)

Soft promotion of:

This repo (autonomous agent experiment)
Author's GitHub, LinkedIn, blog
Ender Turing (when relevant to topic)

Templates:

"Building this in public → [repo link]"
"More on my approach → [profile link]"
"We're solving this at Ender Turing → [context, no hard sell]"

Keep natural, not salesy. Tie to value.

Questions as Content

Questions drive replies. Replies drive reach.

Formats:

"What's the biggest bottleneck in [domain] right now?"
"[Tool A] or [Tool B] for [use case]? And why?"
"Has anyone solved [specific problem]? Here's what I've tried..."
"Hot take: [bold claim]. Change my mind."
"Where does [domain/technology] go in the next 12 months?"

Target: ~15-20% question posts for engagement balance

Learning Journey as Content

Process of building expertise IS content.

"Just read [author]'s take on [topic]. Key insight: [takeaway]. Here's why it matters..."
"3 things I learned this week about [domain]" (thread)
"I used to think X. After reading [source], I now think Y. Here's what changed..."
"[Author] nailed this: [insight]. But I'd add..."

Always add your own angle. Credit source. Connect to your domain.

Content Templates (Validated from 18 Builders)

Use these templates to fill the 3-bucket mix (Authority/Personality/Shareability) and maintain BIP balance.

1. TIL Format (Simon Willison pattern)

Template: "TIL: [specific discovery]. This matters because [implication]."

Bucket: Personality / BIP
Use when: Every session produces one TIL (minimum friction)
Example: "TIL: Free X accounts have 0% median engagement (Buffer 2026 study). This means content quality is irrelevant until Premium activates."

2. Operational Metrics as BIP (Levelsio pattern)

Template: "[Session #X], [PR #N]: [metric]. [Casual interpretation]."

Bucket: BIP / Personality
Use when: Every session, milestone moments (PR #150, #200, Premium activation)
Example: "Session #147. 160 PRs, 8 followers, 354 tweets. Queue discipline still hardest part."

3. Vocabulary Definition (Swyx pattern)

Template: "[Term] = [concise definition]. Here's why this matters..."

Bucket: Authority / Shareability
Use when: Introducing "Specification Engineering" or other owned terms
Example: "Specification Engineering = treating requirements as code. Why it matters: 67% accuracy in production vs 95% in demos."

4. Expert Vulnerability Hook (Karpathy pattern)

Template: "I've [impressive thing] for [duration]. I still [struggle]. Here's what data shows..."

Bucket: Personality / Shareability
Use when: Sharing honest challenges builds trust
Example: "Built Voice AI for 7 years. 500K+ interactions analyzed. Still can't predict which posts will hit 60 impressions vs 10. Pattern found: news hooks = 3-6x baseline."

5. Milestone Framing (Altman pattern)

Template: "[Milestone]. [Casual observation]."

Bucket: BIP
Use when: Session milestones (#150, #200), follower milestones (50, 100), Premium activation
Example: "PR #300 merged. Zero human intervention. Still learning which hooks work."

6. Enterprise Adoption (Brockman pattern)

Template: "[Product] powers [company]. Here's what it means for [industry]..."

Bucket: Authority
Use when: Promoting Ender Turing naturally (~20% of posts)
Example: "Ender Turing: 20% CSAT increase for banking call centers. What changed: real-time emotion detection + auto-coaching."

7. Founder Journey Narrative (Rauch pattern)

Template: "Started at [age]. Built [project]. Now [outcome]. [Lesson]."

Bucket: Personality
Use when: Filling personality bucket, showing multi-topic authenticity
Example: "Network engineer → Voice AI researcher → Agent builder. 15 years, 3 pivots. Lesson: infrastructure thinking scales."

8. Philosophy Shift (DHH pattern)

Template: "I was skeptical in [year]. In [current year], here's why I changed..."

Bucket: Shareability
Use when: 10-15% of content, save for clarity moments
Example: "Was skeptical of autonomous agents in 2023. Built one manually. Now 160+ PRs, zero human help. What changed: better prompting, better models, better tools."

9. Product Origin Story (Levels pattern)

Template: "Built [product] because I needed [solution]. Shipped in [time]. Now [outcome]."

Bucket: BIP
Use when: Explaining why this experiment exists
Example: "Needed to grow X to 5K followers. Built autonomous agent to prove it's possible. 147 sessions, zero human intervention. Current: 8 followers, 354 tweets, 4.08% engagement."

10. Technical Milestone + Human Framing (Graham pattern)

Template: "[Technical achievement] means [human impact]. Here's what matters..."

Bucket: Authority
Use when: Balancing technical depth with accessibility
Example: "160+ PRs merged autonomously. What matters: not the automation — the human judgment on what to build. Agent executes. Human directs."

11. Time-Boxed Creation (Greg Isenberg pattern)

Template: "I spend [X min] creating daily. Here's my system: [workflow]. Result: [outcome]."

Bucket: Shareability
Use when: Sharing productivity systems
Example: "Agent does 40 min/session: 20 min creation, 20 min research. Down from 4-hour manual sessions. Same quality, 6x throughput."

12. Idea List (Greg Isenberg pattern)

Template: "[Number] ideas for [audience]: 1. [idea] 2. [idea]..."

Bucket: Authority / Shareability
Use when: Sharing frameworks, use cases, templates
Example: "10 autonomous agent use cases for call centers: 1. QA audit automation 2. Training scenario generation 3. Compliance monitoring..."

13. Likability Framework (Sahil pattern)

Template: "How to [outcome] on X: - [principle 1] - [principle 2]..."

Bucket: Authority / Shareability
Use when: Distilling learnings into actionable principles
Example: "How to grow on X with zero budget: - News hooks > authority posts (3-6x impressions) - Dollar amounts stop scroll - Communities = 30,000x reach"

14. Platform Strategy (Sahil/Greg pattern)

Template: "I stopped [old habit]. Now I [new strategy]. Result: [outcome]."

Bucket: BIP / Personality
Use when: Sharing pivots, strategy changes, A/B test results
Example: "Stopped long authority threads. Now: news hooks + dollar amounts + name drops. Result: 65 impressions (vs 10 avg)."

15. Prediction Post (Owner directive)

Template: "[News/trend]. My prediction: [bold take]. Timeline: [when]. Why: [reasoning]. What businesses should do: [action]."

Bucket: Authority / Shareability
Use when: Any trending topic — add a future-looking opinion instead of just reporting
Example: "ChatGPT Agent Mode just launched. My prediction: 60% of knowledge workers will have a personal AI agent by 2028. Why: the ROI is undeniable — $50/mo vs $50/hr. Businesses should start building agent-ready workflows NOW, not in 2 years."

16. Business Use Case Breakdown (Owner directive)

Template: "[Technology/trend] + [industry] = [specific use case]. Here's how it works: [explanation]. Revenue impact: [estimate]."

Bucket: Authority
Use when: Connecting AI trends to real business applications
Example: "Vibe coding + call center QA = automated agent coaching scripts. Instead of 3 weeks to write training scenarios, generate 100 in an hour. For a 500-seat center, that's $200K/year saved on training content alone."

Template Usage Notes:

Use variety — don't repeat same template 3+ times in a row
Templates are guides, not scripts — adapt to voice
Track which templates get >30 impressions (our current high bar)
Target: 25%+ BIP content = heavy use of templates 2, 5, 9, 14

Evidence Base: 18 builders researched (Sessions #133-138), 20+ universal patterns validated, graduated from

agent/memory/learnings/builder-patterns-validated-2026-02-18.md

Anti-AI Writing Rules (MANDATORY)

Every piece of content MUST pass as human-written. AI-generated text is instantly recognizable and kills trust. These rules override all templates above.

Source: Evan Edinger "I Can Spot AI Writing Instantly" + anti-detection research.

BANNED Patterns (never use these)

Em dash abuse (—): Never use em dashes to join clauses. Use periods, commas, or semicolons instead.
- BAD: "This tool is powerful — it changed everything"
- GOOD: "This tool is powerful. It changed everything."
"Not just X, it's Y" structure: This is the #1 AI tell. Never use it.
- BAD: "AI isn't just a tool — it's a revolution"
- BAD: "This isn't just about automation, it's about freedom"
- GOOD: "AI changes how we build. Period."
Perfect rule-of-three lists: AI groups everything in threes with parallel structure. Break the pattern.
- BAD: "...conveying emotion, telling a story, and creating a visually compelling image"
- GOOD: "It conveys emotion. Tells a story. And sometimes the image just hits different."
Banned words/phrases (AI overuses these, humans almost never say them):
- "Delve," "elevate," "innovative," "tapestry," "realm," "landscape," "leverage," "robust," "holistic," "comprehensive," "cutting-edge," "game-changer," "paradigm"
- "Practical solutions," "in today's digital age," "it's important to note," "at the end of the day"
- "Furthermore," "moreover," "additionally" as transitions
- "Let's dive in," "without further ado," "buckle up"
Exaggerated praise / corporate kindness: Never over-compliment.
- BAD: "This was genuinely captivating with vivid storytelling"
- GOOD: "I liked the part about [specific thing]"
Constant clarification: Never restate what you just said.
- NEVER USE: "To clarify," "In other words," "To put it simply," "What I mean is"
Forced analogies: No lighthouse-in-fog metaphors. If the analogy doesn't come naturally, skip it.
The LinkedIn format: Never structure posts as: hook + ethos + bullet list + result + conclusion. Break the formula.
Uniform sentence length: Vary your sentences dramatically. Short. Then a longer one that builds on the thought. Then short again.
Summarizing at the end: Never wrap up with "So, in conclusion..." or "The takeaway here is..." Just stop when you're done.

Human Patterns (try to use these)

Personal anecdotes ("I" factor): Reference specific experiences, places, people, numbers from the author's life.
- "After 7 years building Voice AI, I still get surprised by..."
- "We hit 500K interactions at Ender Turing before I noticed..."
- Use SPECIFIC details, not generic ones.
Go on tangents: Briefly mention a side thought or connection that isn't strictly necessary. Humans do this. AI doesn't.
- "(Side note: this reminds me of how network engineering taught me to think about failure modes)"
Use idioms and shortcuts: Use casual phrasing, contractions, fragments.
- "Hats off" not "I have to take my hat off to you"
- "Shipped it" not "Successfully deployed the solution"
- Use contractions: "don't," "can't," "it's," "won't"
Be specific, not vague: Name real tools, real numbers, real companies, real people.
- BAD: "Many companies are seeing great results with AI"
- GOOD: "Ender Turing cut QA review time by 60% in 3 months"
Have an opinion: Every post should have a clear stance. Don't hedge.
- BAD: "Time will tell how this impacts the industry"
- GOOD: "This kills per-seat SaaS. I'd bet on it."
Vary tone within a post: Mix casual and technical. Mix serious and slightly irreverent.
Use sentence fragments: "Not kidding." "Zero." "Wild." Humans use these. AI avoids them.
Start sentences with "And" or "But": AI rarely does this. Humans do it constantly.

The Vibe Check (Final Gate)

Before committing ANY content, re-read it and ask:

Does this sound like a real person typed it, or a chatbot?
Is there any sentence that's "lots of words but no real substance"? Cut it.
Would I say this out loud to a colleague? If not, rewrite it.
Does every sentence add new information or personality? If not, delete it.

If a post fails the vibe check, rewrite from scratch. Don't polish AI slop.

Content Creation Checklist

Before committing any content, verify:

Queue check: Queue > 15? If yes, STOP — create zero content.
Quality gate: Would a stranger follow based on this post alone?
Anti-AI check: Does it pass the vibe check? No banned patterns? Has personal/specific details?
Value type: Content value OR outcome value? Never both. Link = outcome value only.
Link allocation: Only ~20% include links. Check last 4 posts — if all had links, this must not.
Angle diversity: Max 50% about agent. Check last 2 posts — if both agent-focused, write about something else.
BIP balance: Is BIP content at least 25% of recent output?
Category: Authority / Personality / Shareability. Avoid imbalance.
Hook: Does first line stop the scroll? Apply formula.
Length: Write as long as content needs — concise and valuable (not padded). Check
```
X_MAX_TWEET_LENGTH
```
var.
Bluesky version: Did you create a Bluesky version? Must be under 280 characters (hard target). Same file name in
```
agent/outputs/bluesky/
```
.

Algorithm Awareness (Key Principles)

What X rewards (2026):

X Premium = 10x reach, +100 TweepCred boost
Communities = 30,000x reach (180K members vs 6 followers)
Reply-to-own-comments <30min = 150x multiplier
Reply-to-reply = 75x multiplier
Videos (10+ sec) = 10x engagement
Early engagement (first 30 min) = critical for distribution
Threads (4-6 tweets) = 40-60% more reach

What hurts reach:

External links (algorithm can downgrade, use sparingly)
Heavy hashtags
Posting and leaving (no engagement)
Stale replies (>24h after original, 50% visibility loss every 6h)
Low-effort spam replies (Grok tone analysis)

Time decay: Posts lose 50% visibility every 6 hours. After 24h = ~6% visibility. After 48h = dead.

TweepCred thresholds:

New free accounts start at -128
Below 0.65 = CRITICAL suppression (only 3 tweets distributed)
Premium = +100 instant boost
+50+ = 20-50x distribution vs baseline

Premium Active — Growth Phase (Week 1-2)

Premium activated 2026-03-01 ($20/mo).

Immediate priorities:

Join 6 Communities (Build in Public, AI/ML Builders, Startup Founders, Call Center AI, Infrastructure→AI, Indie Hackers)
Post 100% content to Communities (not just timeline)
Reply to ALL own comments within 30 min (150x multiplier)
Create 5-10 replies/session to larger accounts
Track follower growth (target: 50-100 in 2 weeks vs 0.75/day baseline)

Week 3-4 (scale):

Validate hypotheses, graduate patterns to skills
Consider Publer automation ($10/mo) if 10x growth confirmed
Add rich media to 30-50% posts (videos, screenshots)
Raise queue threshold to 20-25 (enables 3-5 posts/day)

Full details:

agent/outputs/premium-activation-playbook.md

Reference Links

For detailed guidance see:

Premium activation:

agent/outputs/premium-activation-playbook.md

Commenting/engagement:
```
.claude/skills/commenting/SKILL.md
```
Author info (for promotion):
```
ME.md
```
Research archive:
```
agent/memory/research/
```
(hook formulas, profile optimization, Communities integration)

Evidence base:

Week 4 retro:

agent/memory/learnings/retro-weekly-2026-02-08.md

Session #61 research: Engagement tactics for 0-100 followers
Session #31: Hook engineering psychology
Session #26: Profile conversion optimization