The Philosophy Behind Writing WITH AI (Not Having AI Write FOR You)
I used to spend 8 hours writing each newsletter.
Now it takes 2 hours.
But here's what most people get wrong about AI writing.
It's not about automation.
It's about collaboration.
The Editor-in-Chief Mindset
When you write with AI, you're not a writer anymore. You're an editor-in-chief.
Think about how magazines work:
The editor decides what stories to run
Writers generate multiple angles
The editor picks the best ones
Writers create drafts
The editor refines them
With AI, you become the editor. AI becomes your writing team.
Your job isn't to type words. It's to make decisions:
What's the core message?
Which headline works best?
What value am I delivering?
Does this sound like me?
Why Claude Code Is a Paradigm Shift
Here's why I use Claude Code instead of ChatGPT or other tools:
Claude Code is an AI agent, not just a chatbot.
With ChatGPT, you prompt once and hope for the best. The AI has to create everything in one shot. The result? Generic, unsatisfying content that sounds like AI.
Claude Code works differently:
It creates a plan and sticks to it
Each phase becomes a separate task
It thinks longer but iterates multiple times
It refines based on your feedback
The output sounds human, not robotic
Think of it like this: ChatGPT is a sprint. Claude Code is a marathon with strategic rest stops.
The Secret: Context Engineering
This is the most important part that nobody talks about.
AI doesn't know what you know. You have to teach it.
Before I write anything, I load my research folder with:
Articles about my topic
Statistics from credible sources
Examples of similar content that worked
My previous newsletters for voice matching
Notes from podcasts or videos
Competitor newsletters for inspiration
This is called "context engineering" - giving AI the raw materials to build from.
Without context, AI writes generic fluff. With context, AI writes like an expert.
I spend 30 minutes gathering context manually (automating this soon). That phase makes the difference between content that gets ignored and content that gets shared.
The 10-Phase Process (Why It Works)
Our newsletter system has 10 phases.
That might seem like a lot, but here's why it's genius:
Phases 1-4: Strategy
Seed: Clarify your idea (with context)
Goals: Define success
Headlines: Generate 20 options
Curate: Pick the winner
Phases 5-7: Structure
Hook: Grab attention
Structure: Map your content
Magical: Choose delivery methods
Phases 8-10: Execution
Outline: Write the draft (using your research)
Iterate: Refine and polish
Final: Production ready
Notice something? Only 30% is actual writing. The rest is thinking, researching, and deciding.
That's the secret. Great writing isn't about beautiful sentences. It's about clear thinking and solid research.
Btw. I'm teaching this exact system plus all my other AI writing frameworks to a small group.
$5/month if you grab one of the last 2 spots at founding price.
Nothing fancy - just me sharing what actually works. You get the tools, I answer your questions, we all write better.
Why Most People Fail with AI Writing
They expect AI to be magic. They type "write me a newsletter about productivity" and expect brilliance.
That's like asking someone who's never met you to write your autobiography.
Instead, you need to:
Feed it real research and sources
Show it examples of your voice
Guide it through a structured process
Make editorial decisions at each step
Iterate, don't accept the first draft
AI handles the heavy lifting. You handle the context and thinking.
The Real Philosophy: Process > Prompts
Everyone obsesses over the perfect prompt. But prompts don't matter if you don't have:
A clear process
Quality sources
Examples to follow
Iterative refinement
That's why Claude Code wins. It follows a process, not just a prompt.
Each phase builds on the previous:
Research informs the seed
The seed shapes the goals
Goals determine headlines
Headlines guide structure
Structure drives the draft
It's systematic. Predictable. Reliable.
What This Means for You
You're not learning to write prompts. You're learning to:
Gather quality context
Think like an editor
Guide an AI agent through a process
Make strategic decisions
Refine iteratively
The Context Engineering Workflow
Here's my exact process:
Before Writing:
Search for 5-10 recent articles on my topic
Copy them into my research folder
Add any relevant statistics or studies
Include 2-3 of my best previous newsletters
Save competitor examples that performed well
Above is the analysis done for my latest newsletter post.
During Writing:
Claude Code reads everything in the research folder
It understands the topic deeply
It matches my voice from examples
It cites real sources and data
It creates something original but grounded
The Result: Content that's informed, accurate, and sounds like me—not like generic AI.
That’s it for today, if you’d like to learn more about writing with AI and get the exact prompts that I use, make sure to join our community!
Luke