Summary: LinkedIn is the professional distribution layer in the cross-posting hierarchy. Posts are condensed versions of blog posts (900–1,500 chars optimal), formatted for scanning with hooks, bold claims, and clear CTAs. Algorithm favors Saves > Comments > Likes; Carousels > Images > Text.

Platform Role

Substack (canonical long-form)
       ↓ condense
LinkedIn Article + Post (professional audience)
       ↓ thread
X/Twitter (real-time, technical audience)

Post Specifications

Spec Value Optimal
Max characters 3,000 900–1,500
Hashtags Unlimited 3–5 niche
Line breaks Critical Double breaks between paragraphs
Images 1–2 1200×627 or 1080×1080
Alt text Required Descriptive + keyword
External links 1 max (in first comment) Canonical URL in comment
Mentions @people/@companies When relevant
Video Native upload 30–90 sec for tech

Algorithm Signals (2024–25)

Signal Weight Strategy
Saves Highest (5× Likes) Create reference-worthy content
Comments High Ask specific questions
Dwell time High Carousels, long reads
Likes Baseline Don't optimize for
Shares Medium Make shareable (insight + credit)

Carousels > Images > Text for reach.

Post Template (from wiki/blogs → posts/linkedin/)

# Headline: [Hook + Key Benefit in ≤ 8 words]

[One-sentence hook: the problem or surprising fact]

---

[2–3 short paragraphs: context → insight → implication]

**Key takeaways:**
1. **Bold claim** — supporting detail
2. **Bold claim** — supporting detail
3. **Bold claim** — supporting detail

[Optional: simple table, diagram, or code snippet]

---

**The reality check:** [Honest limitation or "this won't work when..."]

**Where I'm betting next:** [1–2 concrete future directions]

🔗 **Full deep-dive:** (link in comments)

#Hashtag1 #Hashtag2 #Hashtag3 #Hashtag4 #Hashtag5

Formatting Checklist

  • First line hooks (question, contrarian claim, surprising stat)
  • Double line breaks between EVERY paragraph
  • Bold key phrases for scanning (not markdown **bold** — renders as bold)
  • 3–5 niche hashtags (e.g., #DiffusionPolicy not #AI)
  • CTA: "Full post:" + link OR "What's your experience?" + question
  • No markdown artifacts (raw **bold** renders on LinkedIn)
  • Image alt text if attaching visuals

Hashtag Strategy

Category Examples Count
Niche technical #DiffusionPolicy #FlashAttention #LoRA #WorldModels 2–3
Field #EmbodiedAI #Robotics #MLOps #LLMTraining 1–2
Audience #MLResearch #AIEngineering #PhDLife 0–1

Never: #AI #MachineLearning #DeepLearning — too broad, algorithm ignores.

Cross-Posting from Blog

Blog Section LinkedIn Equivalent
TL;DR Hook + 3 bullets
Introduction Paragraph 1 (context)
Core Idea Paragraph 2 (insight)
Key Insight 1–3 Key takeaways bullets
Technical Deep Dive Optional: 1 diagram/table/code
Limitations "The reality check"
Future Directions "Where I'm betting next"
Further Reading "Full deep-dive:" + canonical URL

Best Posting Times

Window Days Notes
08:00–09:00 local Tue–Thu Morning commute check
12:00–13:00 local Tue–Thu Lunch scroll
17:00–18:00 local Tue–Thu Evening wind-down
Weekend 10:00–11:00 Sat Deep-dive reading time

Consistency > timing: 2–3×/week > perfect timing once.

Engagement Protocol

  1. Post → Reply with canonical URL in first comment
  2. Reply to all comments within 2 hours (signals quality)
  3. Tag relevant people/companies (not spammy)
  4. Share to relevant groups (if allowed)
  5. Save your own post (signals value)

Related Concepts

Sources