16 July 2024

Write Viral LinkedIn & X Posts With Claude 3.5 Sonnet (Tutorial)

2Frameworks

9Insights

Frameworks in this episode

Innovation5 steps

AI Suggests, Human Refines Loop

Use AI for rapid options and human judgment for the publishable result

Read →

Marketing6 steps

Format-and-Style Training Stack

Train AI on an asset’s structure and voice before asking it to create

Read →

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Hot Take· 3

Hot Take02:30

Business Content Has a Platform Problem on X

Kieran argues that weak X performance may reflect the content category rather than faulty writing templates. Marketing and business posts face a difficult environment there, while more outrageous material can attract attention.

A template can work while its topic fails
Platform incentives shape what receives distribution
Marketing content is currently difficult on X
Creators should train on examples that genuinely perform on the target platform

“marketing and business content is rough on X right now”
Kieran Flanagan · 02:30

“if you want to do well on X you should post in just ridiculous things”
Kieran Flanagan · 02:30

#x#distribution#business-content#platform-fit

Hot Take21:30

Claude Is Probably a Better Writer Than Its Critics Think

Kieran challenges the claim that AI only creates average content, arguing that Claude can produce above-average writing when used skillfully. He still rejects the idea that it independently produces finished great writing, assigning that leap to human editing.

Untrained prompting understates model capability
Claude can exceed average writing quality
Great output still requires editing
AI enables capable writers to scale stronger work

“Claude can create above average rri in if you actually know how to use it”
Kieran Flanagan · 21:30

“can it create great rri in no you have to edit it”
Kieran Flanagan · 22:00

#claude#writing-quality#ai-content#editing

Hot Take23:30

LinkedIn’s Algorithm Has Raised the Quality Bar

Kieran says creators can no longer casually riff on LinkedIn and expect reliable engagement. Changes to the platform’s algorithm require more deliberate ideas and stronger execution to earn distribution.

LinkedIn engagement has become harder to obtain
Casual posts are less dependable
Creators need stronger ideas and execution
Platform changes increase the value of preparation

“you can no longer just like Riff on something”
Kieran Flanagan · 23:30

“it's harder to get engagement so you have to really be good”
Kieran Flanagan · 23:30

#linkedin#algorithm#engagement#content-quality

Explainer· 2

Explainer01:00

Why Short-Form Content Is Easier to Build With AI

Kieran explains that short-form posts are a practical starting point because their ideas and structures are simpler to describe than long, deeply argued pieces. He applies the approach to LinkedIn and X and begins extending it to YouTube Shorts.

Short-form assets have more observable structures
LinkedIn, X, and YouTube Shorts are initial use cases
Long-form thought leadership requires more sophisticated development
AI can produce good short-form work before human editing makes it great

“short for Content the structure of the post and the idea are much more simplistic to create than like a long thought for provoking piece”
Kieran Flanagan · 01:00

#short-form#linkedin#x#youtube-shorts

Explainer10:30

AI Can Turn Examples Into Its Own Operating Instructions

Kit highlights a useful capability behind the demonstration: the model can analyze supplied examples and create the instructions it will later follow. The human’s work shifts toward selecting strong evidence and evaluating the resulting guide.

Examples can be converted into reusable instructions
The model can perform much of the analytical documentation work
Human research still determines the quality of the source material
Generated instructions can be reused for later production

“you can have the AI generate its own instructions”
Kit Bodner · 10:30

“you did the research to get the good examples and then you use those examples to get instructions back”
Kit Bodner · 10:30

#instruction-generation#few-shot-learning#examples#automation

Story· 1

Story05:30

The LinkedIn Post That Won by Getting Mocked

The hosts discuss an HR-platform CEO announcing that AI assistants would receive employee-style records. Although commenters ridiculed the post, Kit suggests the controversy may have been intentional because polarized reactions can increase reach.

Provocative positioning can invite predictable criticism
Negative comments may still provide distribution
Dividing opinion can be a deliberate marketing tactic
A mocked claim can still contain valid operational ideas

“do something that actually divides opinion”
Kieran Flanagan · 06:00

“they did it cuz they wanted to get eaten alive”
Kit Bodner · 05:30

#linkedin#controversy#engagement#positioning

Tool· 1

Tool12:30

Add Explicit Exclusions to Control Claude’s Defaults

During the live generation, Kieran notes that Claude tends to add emojis on its first attempt even when that style is unwanted. He explicitly prohibits emojis and hashtags, showing that negative constraints can be as important as positive instructions.

Models may introduce recurring stylistic defaults
State unwanted elements explicitly
Negative constraints reduce cleanup work
Formatting preferences belong in the creative brief

“please don't pleas don't use emojis emojis or hashtags”
Kieran Flanagan · 13:00

#claude#constraints#emojis#hashtags

Takeaway· 2

Takeaway05:00

Stop Treating AI Like an Intern With No Brief

Kit and Kieran compare detailed AI context to onboarding an employee or teaching a business-school student. The model performs better when it receives deep background, examples, and explicit expectations instead of a casual two-sentence request.

AI needs onboarding-style documentation
Detailed context raises output quality
Line-by-line expectations reduce ambiguity
A casual prompt delegates too many decisions to the model

“you're not treating it this like an intern you're treating it like it's a business school student”
Kit Bodner · 05:00

“most people are like oh write me a funny LinkedIn post right it's a two sentence prompt”
Kit Bodner · 06:30

#ai-onboarding#context#prompting#management

Takeaway22:30

AI Scales the Marketer’s Existing Level of Judgment

The episode closes with a distinction between marketers who use AI to improve strong work and those who use it to mass-produce unedited drafts. The hosts frame AI as an amplifier of the operator’s existing standards rather than an equalizer that guarantees excellence.

AI amplifies existing skill and judgment
Great marketers use tools to increase quality and volume
Copy-and-paste workflows scale mediocrity
The operator remains responsible for the final standard

“the great marketeers will use AI to make themselves even better”
Kieran Flanagan · 22:30

“the average marketeers will use AI to copy and paste and make themselves average at scale”
Kieran Flanagan · 23:00

#marketing-skills#scale#quality#human-judgment