The episode challenges the idea that context should be a brief introductory sentence. The hosts describe using a 30-page style guide to create a landing page, arguing that large context windows should be filled with relevant materials when quality matters.

Treat context as a substantive input
Supply style guides and supporting documents when relevant
Use the model's context capacity deliberately
Tailor background information to the requested output

“The context we provided to create the world's best landing page was actually a 30-page style guide, right?”
Kieran · 04:30

“You can add a ton of things to the context window and really tailor those results.”
Kieran · 05:00

#context#prompting#style-guides#landing-pages

Hot Take· 1

Hot Take24:30

Deep Research Should Produce Action, Not Just Reports

Kieran argues that deep research is more valuable when it generates an executable asset rather than stopping at analysis. The same research-to-output approach can create video outlines, blog structures, social threads, or detailed guides for replicating marketing tactics.

Design research around an executable output
Swap in different output templates as needed
Use research to operationalize marketing tactics
Move from analysis toward immediate implementation

“You can actually use that deep research and just swap out the YouTube outline for any outline.”
Kieran · 24:00

“I think of it as a way to actually get very specific on things you can execute on, like take action on.”
Kieran · 25:00

#deep-research#execution#marketing#automation

Explainer· 4

Explainer02:00

The Four Components That Make an AI Prompt Work

The hosts break a strong prompt into four foundational components: a clear goal, a defined return format, explicit warnings, and substantial context. Together, these elements tell the model what outcome to pursue, how to present it, and which failure modes to avoid.

State the desired outcome clearly
Specify the exact return format
Warn the model about likely errors or hallucinations
Provide enough relevant context to tailor the result

“If you want to construct a really good prompt, and this says for the O1 models, but I think this is just good prompt hygiene…”
Kieran · 02:00

“Do I have a goal? Have I stated my return format? Have I give it some sort of warnings to say, like these are the…”
Kieran · 05:00

#prompting#ai#llms#context

Explainer11:00

Give Deep Research the Business Context Behind the Question

Deep research performs better when the model knows who is asking and why. Kieran supplies details about his role, company, customers, technology stack, goals, and challenges so the resulting analysis fits the actual operating environment.

Describe your role and company
Identify relevant customers and industry
Include the technology stack when applicable
State the goals and constraints behind the research

“And so an example would be whenever I am doing something for HubSpot, I give it full details on my role, the company size, the…”
Kieran · 11:00

“And so what I found is if you set the context about who you are, you know, the company you're in, all of the things…”
Kieran · 11:30

#deep-research#business-context#research#ai

Explainer16:00

Why Deep Research Pauses for Clarifying Questions

Deep research commonly asks follow-up questions before beginning a run. The hosts interpret this as a safeguard intended to collect missing requirements and avoid wasting a research run on an underspecified request.

Expect clarification before research begins
Answer missing scope questions directly
Use clarification to reduce reruns
Confirm requirements before launching expensive work

“Deep research now, just so everybody knows, it's always gonna ask clarifying questions.”
Kip · 16:00

“So I think they're trying to reduce the number of runs that you're doing, but it's always gonna ask you a handful of quick clarifying…”
Kip · 16:30

#deep-research#clarification#requirements#workflow

Explainer21:30

Make AI Map Every Video Section to the Runtime

The generated YouTube prompt assigns exact durations to the hook, introduction, context, insights, implications, critical perspective, and call to action. This turns a broad outline into a practical production map for a target runtime.

Set the total video length
Allocate time to each major section
Reserve time for implications and nuance
End with a call to action and engagement loop

“It gives the exact time length of the intro. It gives the exact time to set the context.”
Kieran · 21:30

“So it's given like an exact breakdown map to the 30 minutes, which again is pretty incredible.”
Kieran · 22:00

#youtube#video-planning#storytelling#production

Story· 1

Story18:30

How Kieran Derived His YouTube Outline Style

Kieran created the video-outline style by studying transcripts from videos that had already performed well. He supplied those transcripts as context, then iterated between Gemini and ChatGPT to distill recurring structural patterns.

Select videos with proven performance
Use their transcripts as model context
Ask AI to distill a reusable style
Iterate across multiple models

“this was done using the thing I usually do, which is taking videos that have worked really well, taking the transcripts, adding them into context,…”
Kieran · 18:30

“I did it in Gemini and Chat GPT. I've been switching between both models, just iterating.”
Kieran · 19:00

#youtube#reverse-engineering#transcripts#content-strategy

Tool· 3

Tool05:30

Turn a Proven Prompt Template Into a Custom GPT

Kieran demonstrates a custom GPT that converts a short request into a detailed prompt following a proven structure. This gives novice users a reusable prompt engineer without requiring them to remember every component themselves.

Start with a prompt template that already works
Instruct a custom GPT to reproduce the template
Describe the desired task conversationally
Iterate on the generated prompt before using it

“I'm constantly taking prompts and then turning them into templates and then creating custom GPTs to replicate that template for me.”
Kieran · 05:30

“It will make sure that you give it all of the things it needs, which I think is a good actual checkpoint for people.”
Kieran · 08:30

#custom-gpt#prompt-engineering#templates#automation

Tool09:30

Let Different AI Models Edit Each Other's Prompts

The hosts improve advanced prompts by moving drafts between models such as ChatGPT and Gemini. Each model critiques or extends the other's work, after which the human keeps useful additions and rejects unwanted changes.

Create a detailed draft in one model
Ask another model how to improve it
Move revisions back and forth between models
Keep human control over the final prompt

“But if you flip-flop back and forth in the models, they can edit on each other's work.”
Kieran · 10:30

“It's a combination of two models.”
Kieran · 10:30

#multi-model#gemini#chatgpt#prompt-iteration

Tool17:00

Convert Deep Research Into a Production-Ready YouTube Outline

A custom prompt combines research analysis with narrative strategy and screenwriting to produce a detailed YouTube outline. It searches for statistics, counterintuitive findings, analogies, dilemmas, and quotes before mapping the material into an engagement-focused video structure.

Define the topic, audience, and video length
Research for narrative-worthy evidence
Identify surprising and counterintuitive material
Convert findings into a structured video outline
Include production and discoverability details

“Your mission is to perform a deep multifaceted research on a topic provided and then transform that research into a world-class YouTube video outline.”
Kieran · 17:30

“It's trying to find all of the things in this topic that would actually make it a good YouTube video.”
Kieran · 18:30

#youtube#content-marketing#deep-research#video

Takeaway· 4

Takeaway04:00

Make the Model Show Its Sources and Assumptions

For internal research, Kieran asks the model to cite the documentation and exact location supporting each claim. This makes it easier to verify that outputs rest on real sources rather than hallucinated evidence or opaque assumptions.

Request citations for factual claims
Ask for the exact source location
Inspect the basis for assumptions
Verify that cited documents actually exist

“I'll always say, like, cite the documentation and the exact place where you got the information to make the statement.”
Kieran · 04:00

“So I can see like exactly it's getting it from documents that actually exist.”
Kieran · 04:00

#citations#hallucinations#verification#research

Takeaway09:00

Write the First Prompt Yourself Before Automating

Although AI can accelerate prompt construction, Kieran recommends drafting an initial version yourself. Manual practice builds prompting judgment, while subsequent AI editing can quickly raise a novice's work above average.

Draft an initial prompt manually
Use AI assistance after developing a baseline
Review generated prompts instead of accepting them blindly
Continue editing through conversation

“I don't think you want to rely solely on AIs, but if you are like a novice at Prompton, doing what I just showed you…”
Kieran · 09:00

“And then if you actually play around with it and actually ask it to continue to edit it, okay.”
Kieran · 09:30

#prompting#learning#ai-assistance#judgment

Takeaway13:00

Format Deep Research for the Next AI Task

Rather than requesting a generic information dump, the hosts recommend telling deep research how its findings will be used next. Packaging research for a known downstream prompt reduces rework and can produce a stronger final result.

Define the intended downstream use
Specify the deliverable format in advance
Package findings for the next prompt or model
Avoid generating an unstructured research dump

“And you're way better off to just tell the deep research, it's like, hey, I need this formatted to go with this prompt that I'm…”
Kip · 13:30

“But that's why that's so important.”
Kip · 13:30

#deep-research#workflow#deliverables#prompting

Takeaway24:00

A Great Prompt Is Still a Draft, Not the Finish Line

Even after the generated video outline impresses the hosts, they identify improvements: map the response more tightly to the requested structure and emphasize practical marketing use cases. The broader lesson is to keep refining successful outputs rather than treating them as final.

Review output against the requested outline
Add explicit warnings about unwanted behavior
Strengthen alignment with the target audience
Refine practical use cases before publishing

“But don't just stop there.”
Kieran · 24:00

“The thing I would definitely ask it to ensure it does is map to this outline. Map to the outline and map a little bit…”
Kieran · 24:00

#iteration#quality-control#prompting#content

Frameworks in this episode

Cross-Model Prompt Refinement Loop

Custom GPT Prompt Engineer

Five-Part Deep Research Brief

Goal-Format-Warnings-Context Prompt Anatomy

Marketing Tactic Replication Blueprint

Research-Driven YouTube Outline

Research-to-Output Pipeline

Successful-Example Style Distillation

Insights & moments

⚡Myth Buster· 1

Why One Sentence of Context Is Usually Not Enough

◆Hot Take· 1