LinkedIn AI classroom map
What it is:
- A Skool community created 2025-08-27, with 1,126 members. The courses were built in Aug-Sep 2025, and one locked course was updated 2025-10-02.
- The instructor voice is "Cole" (Nicolas Cole), with example posts from Dickie Bush. It is the same team as Ship 30, and it reuses Ship 30 language: 2-Year Test, 4A, FOR WHO / SO THAT, Sacred Hours.
- Cited outside authorities: Richard van der Blom's annual algorithm report, Chris Donnelly and J.K. Molina.
- It is about 4 years newer than Welsh's LinkedIn OS (about mid-2021) and is built around writing with AI.
Structure (courses → modules)
| Course | Lessons | Access |
|---|---|---|
| (Start Here) Onboarding Checklist | 1 | open |
| LinkedIn AI: Pro Checklist | 1 (long; 7 sections, ~2 weeks each over 8 weeks) | open |
| 1/ LinkedIn Positioning System | 41 | open |
| 2/ LinkedIn Publishing System | 25 | open |
| 3/ LinkedIn Growth System | 41 | open |
| LinkedIn AI Prompt Library | 50 (Position / Publish / Grow) | open |
| 🔒 Social AI Accelerator | 2 | locked (paid: "a private workshop of our LinkedIn strategies going into 2026") |
| 🔒 Social AI Mastermind | 1 | locked (paid: audit of your LinkedIn and business) |
Onboarding: post an intro, then set "Sacred Hours": 2-hour blocks when you are most productive and least reachable, defended like an airplane cabin.
1/ Positioning System (5 modules + bonus)
Six faulty beliefs:
- I'll get fired.
- I'll get too famous.
- People will disagree with me. Answer: disagreement is a feature.
- I posted once and got 3 likes.
- AI use will get me cancelled.
- AI output needs a full rewrite. Answer: lazy prompts give lazy output.
How LinkedIn works. LinkedIn is 3 platforms in one: a career marketplace, a content platform and a business network.
The 3 algorithm keys:
- Relevance, not reach. A "value filter" runs four checks: authority (do you post consistently in your niche), utility, audience match, and quality (spam, hashtags and off-platform links get pushed down).
- Dwell time is the master metric. A comment beats a like, a long comment beats a short one, and long read time beats both.
- Your niche is your network. Posts go to 1st-degree connections first, so connections outside your niche suppress reach.
Biggest mistake: being clever instead of clear. Be clear about the category, the problem, a solution in 5 words or fewer, and the transformation.
Choose your LLM.
- Claude is the recommended default for writing: best voice matching (feed it 3-5 of your best posts), with Projects to hold pillars and brand rules. ChatGPT is for images and deep research. Gemini only if you already live in Google Workspace.
- Cole's own stack: Claude + ChatGPT + Gamma.
- Assignment: pay for one tool and use only it for 30 days.
Position yourself: the 4-step "Information Advantage" method.
- Get very specific about your experience, skills and interests, and look for unusual combinations. This is the 2-Year Test again.
- Run an AI interview that acts as an "Information Advantage mirror". It returns 5 target niches and 3 content pillars.
- Turn that into an industry niche plus one line: "I help {audience} with {service} to {outcome}." A Positioning Consultant prompt generates 30 micro-niches, 20 adjacent markets and headline variants.
- Treat the profile as a landing page: WHO / WHAT problem / SO THAT.
Optional credibility types: time, a number, an industry accomplishment, or a statement of interest ("Obsessed with {topic}").
Profile.
- Photo: current, face and eyes visible. Banner: industry-relevant, text kept in the center safe zone.
- Headline: the one-liner plus optional credibility.
- About, in 4 parts: why believe you → how you do it → what they get → who you are. The prompt writes 3-7 variants of 250-300 words.
- Verify through CLEAR, add the custom website button and turn on Creator Mode. Skip Premium at first.
- Featured section: newsletter or lead magnet first.
- Bonus: visual brand. "Make decisions once":
- 2-3 colors (dark background + one accent).
- One sans-serif font.
- 2-3 layout templates on a 1200×1200 canvas.
- One consistent screenshot style.
2/ Publishing System (4 modules)
- Five keys.
Write to one reader.
Formatting: patterns 1-3-1, 1-1-3-1 and 1-5-1. One-line sentences, paragraphs of 2 lines at most, and never break a sentence across lines.
Post 3-5 times a week (they post once a day), and every post carries media.
- Mix: 2-3 short posts plus 1-2 long posts a week.
- Short-form is ~25-150 words; long-form is ~200-450 words.
Use 5 proven formats:
- long-form + media
- carousel
- tweet screenshot
- short video + long text
- handwritten infographic
Carousels get the most consistent distribution, and video is rising.
Review metrics once a week (Saturday, ~30 min): impressions, likes, reposts, comments and new followers. Write 5 variations of each winning post.
- Voice archetypes. These are the 4A again: Actionable / Analytical / Aspirational / Anthropological. Two tone modifiers can sit on top: the Opinionator and the F-Bomber. Pick one primary voice and one secondary. A 5-round game prompt finds yours.
- Lean Writing Method.
- 3 content buckets: Futurism, Your Story, Actionable Advice. News is discouraged.
- 3-5 pillars, each run through all 3 buckets.
- 10 "magical" expanders: Tips, Stats, Steps, Lessons, Benefits, Reasons, Mistakes, Examples, Questions, Personal Stories.
- Then multiply by audience archetypes and sub-topics. They claim 3×3×10×3×5 = 1,350 ideas.
- "Outlines as content": post the bare list as a short post first. If it gets traction, expand it.
- Expansion recipe:
- Hook in the first ~200 characters, one of 5 hook types.
- Intro in Problem-Agitate-Solve form.
- Sections, each with a header that is a full sentence, a 1-line opener, 3-5 tactical points and a 1-line takeaway. The headers carry ~80% of the value.
- CTA as a question. Links go in the first comment.
- Content calendar.
- Rotate the three buckets every day.
- Optional weekday themes:
- Mon: case study
- Tue: industry opinion
- Wed: actionable listicle
- Thu: curation
- Fri: what I did this week
- Sat: random idea with your twist
- Sun: vulnerable lesson
- Post at 8:30 AM ET. Follow the plan exactly for 4-6 weeks, then adapt.
- Batching: one pillar per session (1-2 hrs), outlines across all 10 expanders, then schedule them a week or more apart. Each week, expand the best outline into a long post.
- Use LinkedIn's native scheduler. They suspect third-party schedulers get less reach.
3/ Growth System (7 modules + bonuses)
Text hooks.
- 3-line rule: only ~125-150 characters show before "see more". Line patterns: 1-0-1, 1-1-1 or 1-1-0.
- 10 contrast patterns: small/big, hard/easy, far/near, few/many, hidden/obvious, wrong/right, expectation/reality, now/later, scarcity/abundance, common/rare.
- Reverse-engineer viral hooks into fill-in templates with an AI prompt.
- "Interview technique": AI turns a viral post into questions, you dictate the answers, then AI rebuilds the post in your words.
- The visual hook must match the text hook.
Carousels.
- Slide 1: hook. Slide 2: re-hook. Then one point per slide. Last slide: CTA. The examples run about 5-16 slides, with a mid-roll CTA in long ones.
- Built in Gamma (Social format, "preserve this exact text", "---" for slide breaks, branded theme, ~1 image per card, export as PDF). The internal SOP repurposes threads and newsletters each week.
Photos.
- 6 styles: behind the scenes, transformation, juxtaposition, pets, little-known credible people, milestones.
- Screenshots as proof ("Tweet Pic").
- "Photos as sawdust": a quarterly photo day, plus an AI avatar (HeyGen or ChatGPT).
Infographics. 5 types: framework, data, comparison, checklist, hand-drawn. Tools: Napkin.ai (easy), Canva/Gamma, or Figma Make + html.to.design (advanced); hand-drawn via Fiverr or HandTextAI.
Video. Only after 6+ months of proven writing, and education rather than entertainment. Two types: shorts of 30-120s and long-form of ~10 min. Three methods:
- cut clips from long video
- read a proven post on camera
- a faceless AI short (Descript)
Also a CapCut 60-second template and a recording checklist (Loom, CapCut, light placement, mic tiers).
Connection engine.
- Welcome DM within 24h with your 3 best posts.
- Use the full 200 connection requests a week, aimed at adjacent creators' audiences and 2nd-degree contacts.
- Connect with people who engaged or viewed your profile, 5-10 a day.
- A weekly "Authority Library" curation of 10-12 examples, compounding into big compilations.
- "Comments as consulting": ~30 min a day, answer every comment on your own posts, and post a bonus comment after 1h.
AI comment drafts are allowed only if you rewrite them by hand. No automation tools: LinkedIn penalizes them, and the team does it all manually.
Daily checklist:
- Morning, 10 min: viewers, 3-5 requests, replies.
- During the day: 3-5 thoughtful comments.
- Evening, 5 min: welcome DMs.
Conversion system.
- Profile CTAs (Premium button, a line in the About section).
- Autoplugs on ~25% of posts. Placement: first comment, or added to the body ~2h after posting, or a mid-carousel slide.
- 5 opt-in CTA templates: Classic / Credibility / Benefits / Pivot / Proven Path.
- "Pitch in Public": publicly critique a prospect's gap and tag them. This is a sales tactic.
- Bonuses.
- Viral Drop: a free asset traded for a comment keyword, delivered by DM, with a "DMs are broken" follow-up after 24h. A list-building tactic.
- Template Vault: 16 hook templates (If/Then, Brutal Story, Average Person Is Wrong, Stop/Start, Misconception Reframe...).
Prompt Library (50 prompts)
- Position: info advantage, bio, banner, About, skills, brand identity, lead magnet.
- Publish: voice archetype, AI Voice Profile, audience research, archetypes, the Content Ideas mega-prompt, pillars, bucket matrix, sub-topics, 10-way expander, outlines, outline to long-form (≤3,000 characters), 10-day quick start, 30-day calendar.
- Post generators, all 200-300 words:
- Story (hook of 9 words or fewer)
- Quick Listicle
- Hot Take
- Case Study
- X vs Y
- Grow: hook optimizer, contrast hooks, comment assistant, CTA plug, DM assistant, viral drop and others.
Formats taught
- Short text post: ~25-150 words, or 1-3 sentences, or 3-5 bullets. Used to test topics, including "outlines as content".
- Long text post: ~200-450 words; hook (~200 characters) → PAS intro → full-sentence headers → CTA question.
- Carousel: hook slide / re-hook / one point per slide / CTA slide, built as a Gamma PDF.
- Tweet screenshot, photo post, infographic (5 types), short video + long text.
- Post-generator shapes: Story, Listicle ("After {time} of {experience}, here's {N} things..."), Hot Take, Case Study, X vs Y. All 200-300 words.
- Profile: headline "I help {audience} with {service} to {outcome}" + credibility; About in 4 parts, 250-300 words; banner in the safe zone.
Cadence: 3-5 posts a week (once a day at most), at 8:30 AM ET; metrics weekly; batching weekly. Engagement: 15-20 min a day; up to 200 connection requests a week.
vs Welsh LinkedIn OS
Compared against [[06-reference/2026-10-02-welsh-linkedin-os-applied]].
What is newer or different:
- Algorithm model. Welsh optimizes the hook and posting daily. LinkedIn AI centers on dwell time + relevance + network fit. This matches the 2026 evidence in the Welsh note (LLM-based ranking, topical consistency). Welsh's dead tactics (hashtags, engagement partners) are gone here; this course names spam and hashtags as things that get pushed down.
- Cadence. Welsh says 5-7 posts a week. This course says 3-5 a week with a short/long mix, which is closer to our 2-posts-a-week plan.
- Carousels and media on every post. Welsh is text-first. This course says carousels get the best distribution, which fits the vendor data in the Welsh note.
- AI in the loop. It recommends Claude by name, a voice profile built from 3 samples, and an interview-then-rewrite method. Welsh predates LLMs.
- "Outlines as content." A cheap way to test a topic before spending an hour on a long post. Welsh repurposes winners only after 60-90 posts.
- Same ideas as Welsh, under new names: the headline formula ("I help {who}..." in both), the topic × style idea matrix (now buckets × pillars × expanders), comment-heavy engagement and "never pitch cold".
- Still risky: the first-comment link trick (Welsh's note cites a vendor saying it is now penalized too; this course still recommends it), Viral Drop comment-gating (engagement bait), and Pitch in Public (client acquisition).
What to use (Ray-selected, awaiting his read):
- The 3-line hook rule + 10 contrast patterns as the hook checklist for the 2 posts a week.
- Short/long split: cut each site piece into one short post (~100 words, an outline-as-content test) and one long post (~300 words, full-sentence headers). This replaces Welsh's daily cadence.
- One carousel a month, built with Gamma or our own template. This upgrades the monthly "document post" in the Welsh 5-hr plan.
- AI Voice Profile from 3 of Ben's real pieces. Ray already holds [[VOICE]]/[[STYLE]], so check those against the prompt's output, don't replace them.
- Comments-as-consulting and the daily checklist, scoped to the ~15 named HCLS prospects (Welsh's outbound list). Skip the 200-requests-a-week volume.
Skip: Viral Drop, Pitch in Public, the autoplug and opt-in CTA machinery, the lead-magnet system, the Social AI upsells. These are services, sales or list-selling mechanics. They conflict with the phData rule (personal brand only; no clients, no selling services) and with our goal (buyer credibility for phData, not audience monetization).
Related
- [[06-reference/2026-10-02-welsh-linkedin-os-applied]]
- [[01-projects/personal-brand/ship30/2026-10-05-ship30-curriculum-map]]
- [[01-projects/personal-brand/2026-09-29-nov30-credibility-targets]]