Kieran argues that AI-generated templates are tools for reverse-engineering expert practice rather than blindly copying finished work. A user can extract the structure behind an expert example, turn it into reusable instructions, and then adapt those instructions to personal taste and style.

Study experts with clearly demonstrated craft
Extract the structure behind why their work succeeds
Ask AI to express that structure as reusable instructions
Edit the resulting template to make it your own

“It is not a copy and paste tool.”
Kieran · 11:00

“It is an accelerated learning tool.”
Kieran · 11:00

#templates#learning#reverse-engineering#creativity

Hot Take· 2

Hot Take08:00

Your AI Results Reflect How Well You Manage the Assistant

The hosts challenge the idea that poor AI output always demonstrates a model limitation. Treating AI like an employee means supplying context, defining the role, documenting the craft, and showing concrete examples of successful work.

A vague instruction produces a weak result
Domain expertise must be transferred to the assistant
Context and examples define the expected standard
Skilled practitioners can use AI to replicate their craft

“Usually your results with AI are reflection of you as a manager of how you are managing that AI assistant.”
Kieran · 08:30

“If you do not set context, if you do not teach the AI how to do a role, then it is not going to be…”
Kieran · 08:30

#management#domain-expertise#context#ai-assistants

Hot Take26:00

Look for Problems AI Has Downgraded From Impossible to Hard

The closing idea, attributed to Microsoft CTO Kevin Scott, is that each generation of AI converts some previously impossible tasks into merely difficult ones. The hosts argue that taking on a consequential hard problem can create differentiation because most competitors will still avoid it.

Reassess formerly impossible projects after model improvements
Identify the newly feasible problem with the greatest impact
Accept that meaningful opportunities may remain difficult
Use difficult execution as a competitive moat

“the best way to think about AI is it makes the impossible problems just really hard.”
Kieran · 26:00

“Because most people will not do the hard thing.”
Kieran · 26:00

#innovation#strategy#competitive-advantage#ai

Explainer· 3

Explainer17:30

Three Types of Short-Form Content That Earn Attention

Kieran identifies three short-form content categories: a focused educational lesson, a counterintuitive spicy take, and a resonant “head nod” observation. The head-nod format articulates something audiences already sense but have not expressed clearly themselves.

Teach one concise lesson rather than several
Use counterintuitive arguments for a spicy take
Articulate an unspoken shared belief for a head-nod post
Categorize source material before choosing the final post style

“There are three types of short-from content that do really well on the internet.”
Kieran · 17:30

“The head nod is everyone kind of believes this, but no one has been able to articulate it.”
Kieran · 18:00

#short-form-content#linkedin#content-strategy#social-media

Explainer16:30

When a Proven Prompt Should Become an AI Agent

Once a prompt reliably performs a defined task, the hosts suggest integrating it into an agent for faster and more repeatable execution. Kieran's LinkedIn agent categorizes transcript material, combines it with a post style, and creates a first draft as part of his content process.

Start by proving the task with a prompt
Encode learned categories and templates into the agent
Use the agent for a narrow, repeatable process
Keep humans involved in the broader team workflow

“Agent is basically when I've actually figured out how to do something via a prompt.”
Kieran · 17:00

“I believe in the future, every single team is going to be made up of humans and virtual agents.”
Kieran · 19:00

#agents#automation#workflows#linkedin

Explainer22:30

Graduate From a Narrow Agent to an App When Complexity Grows

The hosts distinguish agents that automate specific processes from apps that coordinate multiple data sources and more complex behavior. AI coding tools such as Replit's agent can make app creation accessible to people without traditional software-development backgrounds.

Use an agent for a specific automated process
Consider an app when several data sources are involved
Use natural-language coding tools to prototype ideas
Expect the AI to propose technical implementation choices

“You kind of graduate to needing an app when you need lots of different data sources.”
24:00

“What you're trying to do gets very complex.”
24:00

#ai-coding#replit#apps#agents#prototyping

Story· 1

Story05:30

Why Klarna's AI Needed Better Human Onboarding Documents

Klarna's first attempt to introduce AI into customer support reportedly struggled because its onboarding documentation was inadequate. Rewriting the documentation for human representatives also gave the AI clearer examples of how the work should be performed.

The initial AI deployment was not successful
The model itself was not identified as the root problem
Weak onboarding documents prevented clear instruction
Better documentation improved both human and AI execution

“They did not have great onboarding docs.”
Kieran · 06:00

“Once they rewrote the onboarding docs for the human, the AI agents all got better because now they understood how to do that task and…”
Kieran · 06:00

#klarna#documentation#customer-support#onboarding

Tool· 2

Tool01:30

Let Claude Write Your First Prompt

The hosts recommend asking Claude to construct a strong initial prompt instead of laboring over one from scratch. Users should clearly describe the desired outcome, provide relevant supporting material, and then refine the generated prompt through conversation.

Describe the result you want in clear language
Ask Claude to generate the initial prompt
Include supporting documents that clarify the task
Iterate several times before judging the output

“Claude is a much better prompt engineer than most of us.”
Kieran · 01:30

“With AI and with prompting, you're iterating a lot of times.”
03:30

#prompting#claude#ai-tools#iteration

Tool13:30

Turn Expert Content Banks Into Reusable Style Templates

YouTube transcripts, SlideShare presentations, LinkedIn posts, and X threads can serve as source material for studying expert styles. By grouping creators by style and supplying representative examples to Claude, users can generate templates that reproduce selected characteristics at scale.

Collect a substantial bank of expert content
Group creators or examples by identifiable style
Use transcripts and posts as model context
Generate a template for each style
Choose and test styles according to the task

“YouTube is a treasure trove of content.”
Kieran · 14:00

“you can categorize all of those creators into different styles and then teach Claude what that style is.”
Kieran · 14:00

#content#claude#style#templates#creators

Takeaway· 2

Takeaway04:00

Four Prompting Rules for OpenAI's O1 Model

The episode summarizes OpenAI's guidance for prompting its O1 preview model. Effective requests are simple and direct, avoid unnecessary chain-of-reasoning instructions, separate sections with delimiters, and include only relevant context.

Make requests simple and direct
Do not request explicit chain-of-reasoning steps
Use delimiters to separate prompt sections
Remove context that the model does not need

“The first is be simple and direct in your asks.”
Kieran · 04:30

“And the fourth is basically only give it the context it needs.”
Kieran · 05:00

#openai#o1#prompting#context

Takeaway14:30

Test Sales Outreach by Writing Style, Not Just Traditional Metrics

Generative AI makes it practical to produce outreach emails in several consistent styles, including witty, direct, informational, and controversial variants. Teams can compare which style actually produces meetings instead of relying on subjective beliefs about customer preferences.

Create several clearly differentiated writing styles
Apply each style consistently to outreach emails
Measure which variants generate booked meetings
Replace subjective tone debates with structured testing

“I can have a range of different writing styles that Jen AI can use to create outreach emails, whether they're witty, whether they're direct, whether…”
Kieran · 14:30

“Now you can just test, right?”
15:30

#sales#email#experimentation#writing-style

Frameworks in this episode

Agent-or-App Complexity Threshold

AI-Accelerated Expert Templating

Five-Step AI Expertise Ladder

Four-Part Direct Prompt Checklist

Impossible-to-Hard Moat Filter

Prompt-to-Agent Graduation Rule

Short-Form Content Resonance Triad

Structured Style Experimentation

Teach the AI What Good Looks Like

Insights & moments

⚡Myth Buster· 1

AI Templates Can Accelerate Learning Without Becoming Copy-and-Paste

◆Hot Take· 2