Welcome to AI Maker Lab's first post 👋🏻

My free posts show you what's possible with AI. This post shows you exactly how to build it.

This is the complete system I used to build AI that grew my newsletter from 0 to 4K subscribers in 4 months—the exact blueprints, prompts, and frameworks I've never shared publicly.

Everything I wish I'd known months ago about turning Claude (or ChatGPT) into a thinking partner that actually knows your work instead of just another AI tool.

Let's dive deep.

You open Claude for the third time today and face the same choice: explain your context from scratch, or hope the documents you uploaded to Project Knowledge actually help.

Maybe you’re starting fresh with Claude.

Maybe you’ve been experimenting with uploads and custom prompts for weeks but still getting generic responses.

Maybe you’ve uploaded the same strategy documents three times hoping this conversation will be different.

Either way, you end up in the same exhausting place:

“I’m working on a marketing campaign for B2B SaaS companies focused on productivity tools. Our target audience is operations managers at mid-sized companies who are overwhelmed by tool sprawl but skeptical of new solutions. We’ve tried content marketing but engagement is low. The CEO wants to pivot to thought leadership but I’m not sure that addresses the real problem—we’re burning budget on approaches that aren’t working and I can’t afford another failed quarter...”

Five minutes later, you get frameworks that could apply to any B2B company. Generic advice that ignores the three failed campaigns you can’t repeat. Surface-level suggestions that don’t account for your CEO’s communication style, your budget constraints, or the fact that you’ve already tested the obvious approaches.

The AI doesn’t remember that you tested three different content approaches last month. It doesn’t know your audience responds better to case studies than theoretical frameworks. It has no context about your company’s positioning, your CEO’s communication style, or the quarterly pressure you’re actually facing—even though you’ve uploaded documents covering all of this.

Here's the AI system that ended this cycle for me

Instead of starting every conversation from scratch, I built what I call a "persistent intelligence system" - AI that understands my work the way a senior colleague would after six months of working together. Claude knows my communication style without explanation, understands my strategic priorities without briefing sessions, and references past projects when helping with current decisions.

My newsletter grew from 0 to 6K+ subscribers in 6 months because I had an AI partner that could analyze my actual audience data, reference my content performance patterns, and suggest strategies based on what specifically works for my readers - not generic growth advice.

My newsletter intelligence system has two components working as one brain:

1. The Master Prompt serves as the the brain of the system. It defines who I am professionally, what I'm trying to achieve, and how I want Claude to approach every conversation about my work. This stays consistent and shapes how Claude thinks about any request I make.

2. The Knowledge Base serves as the memory bank. It contains my past newsletter content, performance data, reader feedback, growth strategies, and strategic frameworks. This grows over time as I add new content and learn what works.

When I ask Claude about content strategy, it doesn't just pull from general marketing knowledge. It references my specific audience data, analyzes patterns from my highest-performing posts, and suggests approaches that align with my unique voice and positioning.

Here's exactly how it works:

• Instead of: "What should I write about this week?" → Generic AI topic suggestions

• I ask: "Based on my content performance data and recent reader comments, what newsletter topics would drive both authority positioning and subscriber growth?"

• Claude responds by: Analyzing my highest-performing posts, identifying patterns in reader engagement, and suggesting topics that build on what's already working while addressing gaps in my content coverage.

The system turns Claude from a generic assistant into something that actually knows my business—like having a colleague who remembers everything we've worked on and gives better advice each time we talk.

Why Claude works best for this system

I tested both Claude Projects and ChatGPT Projects for three months before committing to this system. The biggest difference: how much of your knowledge they keep “in active memory.”

• Claude Projects: It offers a very large context and lets you attach more than 50 sources as long as the total fits the context limit. Practically, this means Claude can reference much more of your material within a single conversation.

• ChatGPT Projects: Uses RAG (Retrieval Augmented Generation) to search through uploaded files and pull relevant snippets when it thinks they’re needed. You’re limited to 25 files.

Here’s what this means in practice

When I ask Claude about newsletter strategy, it simultaneously considers my content performance data, past reader feedback, my positioning documents, and my writing style guide—because all of it is loaded in context. It connects patterns across my entire knowledge base.

When I tested the same setup in ChatGPT Projects, it would answer questions based on whichever file snippet it retrieved. Ask about growth strategy? It pulls from one document. Ask about content voice? It retrieves from another. It never synthesized across my full knowledge base because it never had access to everything at once.

Here’s how the comparison looks like for Claude vs ChatGPT

Here’s the conclusion

• Choose Claude Projects if: You work with extensive documentation, need accurate research and analysis across your full knowledge base, or require transparency about how much context you’re using.

• Choose ChatGPT Projects if: You need built-in image generation or want to pull from past chat history as part of your knowledge base.

For a comprehensive persistent intelligence system like the one I’m writing you to build, Claude’s larger context window, unlimited file capacity, and full-context approach make it the only practical choice. You need the AI to synthesize across everything you’ve uploaded, not just retrieve fragments.

Three more reasons Claude dominates this system

There are three other reasons this guide works best in Claude's ecosystem:

• Artifacts: Claude Sonnet 4.5 excels at generating functional apps and dashboards with minimal errors. When I ask it to visualize my newsletter performance data, I get working tools, not buggy code that needs debugging.

• Google Drive Integration: All my projects and notes live in Google products. Claude's direct integration means I can reference files in Project Knowledge without manual uploads or copy-pasting.

• MCP (Model Context Protocol): This connects Claude directly to Notion, Linear, Todoist, Apple Notes, WhatsApp, and other tools I use every day. (I'll be diving deep into MCP workflows in upcoming posts - this integration is game-changing.)

You can adapt this framework for ChatGPT or other platforms, but these integrations make Claude the natural choice for persistent intelligence systems.

Building your persistent intelligence system

You're going to build your version of this system by following my exact blueprint, then adapting it for your work domain. By the end of this section, you'll have a working master prompt and knowledge base that transforms how Claude understands your professional context.

Step 1: Master prompt construction

The master prompt is the brain of your system. It defines how Claude should think about your work, what operating modes to use, and how to approach different types of requests. Here's my complete master prompt that turned generic Claude into my newsletter growth strategist: