Why Your Writing Breakthroughs Keep Disappearing

How I let AI analyze my writing to fix chat chaos.

I've been building Claude Project systems for months, but there's always been one painful step: creating the initial project setup from scratch.

Whether you're using Claude or ChatGPT Projects for the first time or building sophisticated business intelligence systems, the initial setup process is brutal. Defining your voice, articulating your audience, crafting those first project instructions—it typically requires hours of manual work before you can start getting value from persistent conversations.

Greg Wolford shared his technique that eliminates this bottleneck entirely. Instead of spending hours defining your voice and audience manually, you upload five recent posts and let Claude/ChatGPT extract the patterns. It identifies your unique positioning, communication style, and audience concerns—often surfacing insights you hadn't consciously articulated. This AI-analyzed foundation becomes the base layer that makes everything else work better.

His three-layer system takes 20 minutes to set up and eliminates the hunt through old conversations entirely. If you're already using Claude/ChatGPT Projects but dreading the setup process, or if you want to see how AI can bootstrap its own understanding of your work, this solves the exact problem you're facing.

Greg Wolford writes about using AI tools to enhance rather than replace human creativity. His craft-first approach helps writers integrate AI thoughtfully while preserving their authentic voice. He's also the author of "The Maker's Heir," a literary sci-fi novel about AI, memory, and what it means to care. Find more practical AI writing guidance by subscribing to his newsletter gregwolford.substack.com

Before we hand it over to Greg, you might want to check out his latest posts.

1. What will writers still do better than AI in ten years?

2. The Death of the Writer's Block Excuse: When AI Makes Everything Possible (And Why That's Terrifying)

3. The Future-Proof Writer: Building Skills That AI Can't Replace

Greg, take it away.

---

Hello, it's me Greg 👋🏻

I found a way to stop losing my best AI insights, and it starts with something that feels backward: letting the AI read my old posts and tell ME what my voice and audience are.

Last week, I had a breakthrough conversation with ChatGPT about my audience's pain points. The AI helped me connect my personal story to my readers' struggles in a way that made everything click about my content strategy.

Today I needed that insight for a post I'm working on. The problem is that I have no idea which of my 200+ ChatGPT conversations it's buried in. The titles tell me nothing—"Content strategy," "Writing advice," "Blog ideas." All useless.

After 20 minutes of hunting through conversations, I gave up and recreated the insight from scratch.

Sound familiar? Here's what I discovered works.

The Fix: Three-Layer System That Actually Works

Claude and ChatGPT have Projects. These projects let you upload files that persist across all conversations. Projects also include a project description and project instructions. Every chat within the project has access to these files and instructions.

Instead of treating each AI conversation as a separate event, I created a project system with three interconnected layers.

Layer 1: AI-Analyzed Project Setup

The trick: I uploaded five recent Substack posts to Claude and said, "Based on these examples of my writing, create a project description that captures my audience, topics, and approach. Then write project instructions that reflect my preferred style and voice."

What I got back:

AI-Generated Project Description: "Writing guidance and philosophy for authors using AI tools in their creative process. Focus on craft-first approaches that enhance rather than replace human creativity. Target audience: fiction and nonfiction writers who want to integrate AI thoughtfully without losing their authentic voice or creative control."

AI-Generated Project Instructions: "Emphasize practical, tested approaches over theoretical advice. Use personal examples from actual writing projects. Address common concerns about AI undermining creativity by showing how it can support craft development. Keep tone conversational but authoritative—someone who's done the work, not just theorizing about it. Always include specific prompts, workflows, or actionable steps readers can implement immediately."

The AI recognized patterns in my writing that I hadn't articulated. It identified my craft-first philosophy, my focus on practical implementation, and my audience's concerns about AI and creativity.

Now every conversation starts with that context loaded. No explaining my audience again. No rehashing my approach. We jump straight into advanced creative work.

How to Set This Up

• For blog writers: Upload your five most recent posts

• For newsletter writers: Upload your last 3-5 newsletters

• For book writers: Upload your outline, character sketches, or sample chapters

• For freelancers: Upload samples from your best client work

Then use this prompt:

"Analyze these writing samples and create two things:

1. A project description that captures my target audience, main topics, and overall approach.

2. Project instructions that reflect my voice, style preferences, and how I like to work with content. Make both specific and actionable."

The AI will analyze patterns in your work—word choice, sentence structure, topics, audience references, and examples you use—and distill them into clear project guidelines.

Layer 2: The Living Project Memory Document

Besides the AI generated project settings, I have one or more evolving documents that capture my project's growing knowledge base.

What goes in the document:

• Key insights about AI and writing craft

• Reader questions and concerns that keep coming up

• Successful workflows and techniques

• Examples and case studies

• New tools and approaches to explore

An Example "AI Writing Craft Memory.md" snippet:

# AI Writing Craft - Project Memory

## Core Philosophy
**Craft-First Approach** - AI enhances human creativity rather than replacing it. Writers maintain creative control while using AI to amplify their natural abilities.

From 8/23 conversation: readers worry most about losing their voice, not about AI taking over. Focus on showing how AI preserves and develops authentic voice.

**Practical Over Theoretical** - Readers want workflows they can use today, not abstract discussions about AI's future.

## Best Insights So Far
8/25 - Reader research workflows: NotebookLM transforms scattered research into story-ready insights
8/23 - Memory persistence: Claude Projects solve the "lost breakthrough" problem most writers face
8/20 - Voice preservation: AI collaborates rather than generates when used properly

## Current Challenges
How do writers overcome the "blank page" feeling when starting AI conversations? What prompts work best for different types of creative problems?

Combined with the AI-generated project description and instructions, this creates a complete knowledge base that every conversation can access and build upon.

Layer 3: Persistent Conversations

Every new conversation in the project automatically has access to both the AI-generated setup AND the accumulated memory document. Nothing gets lost. Everything builds.

Get to Work Faster

Setting up projects like this allows me to jump in and write. The AI knows what I’m doing, my goals, and my priorities.

Before Projects:

• Started every conversation explaining my philosophy about AI and writing

• Lost track of reader questions and successful techniques I'd discovered

• Repeated the same conversations about craft-first approaches

• Felt like every AI session began from zero

After Projects:

• Jump straight into developing specific writing techniques

• Build sophisticated frameworks over multiple sessions

• Never repeat basic context about my audience or approach

• Feel like I'm working with a creative partner who understands my philosophy

For example, when I open a new conversation in my "AI Writing Craft" project, I type, "I want to explore how writers can use AI for character development without losing creative control."

Claude immediately understands:

• My craft-first philosophy (from AI-generated project instructions)

• My target audience of writers concerned about authenticity (from AI-generated project description)

• Previous insights about voice preservation and practical workflows (from project memory)

• My preference for actionable techniques over theory (from analysis of my writing samples)

No explanation needed. We jump straight in.

Your Project Setup Guide

You can set up a project in less than twenty minutes:

Step 1: Gather Your Writing Samples

• Blog writers: Your 5 most recent posts

• Newsletter writers: Your last 3-5 issues

• Book writers: Outline, character sketches, or sample chapters

• Content creators: Your best-performing pieces across platforms

Step 2: Let AI Create Your Project Foundation

1. Open Claude and click "New Project"

2. Name it after your writing project

3. Upload your writing samples

4. Use this prompt: "Analyze these writing samples and create two things: 1) A project description that captures my target audience, main topics, and overall approach. 2) Project instructions that reflect my voice, style preferences, and how I like to work with content. Make both specific and actionable."

5. Copy the AI's responses into your project description and instructions fields

Step 3: Upload Your Project Memory Document

Create a text file called "Project Memory.md" with this template:

# [Your Project Title] - Project Memory

## Core Philosophy/Themes
**[Key Theme Name]** - [Brief description and why it matters to your audience]

## Best Insights So Far
[Date] - [Topic]: [Key discovery]

## Current Challenges
- [What you're working on solving]
- [Questions you're exploring]

## What Works
- [Successful approaches or examples]

Upload this file to your project.

Step 4: Your Enhanced Session Protocol

1. Start conversations by diving straight into specific work (no context explanation needed)

2. End productive sessions by updating your Project Memory document

3. Refine AI-generated project instructions as you discover what works best

Step 5: The Power of AI-Analyzed Context

AI-Generated Description + AI-Generated Instructions + Project Memory = Complete Context

Instead of guessing what your voice and audience are:

• Let AI analyze your actual work

• Extract patterns you might not see

• Create precise guidelines based on your proven approach

• Build on that foundation with accumulated insights

Why This Three-Layer System Changes Everything

Now every conversation starts with complete context. When I type "How do I help writers overcome creative blocks when using AI for character development?" Claude immediately understands:

• My craft-first philosophy (from AI-generated project instructions)

• My target audience of writers concerned about authenticity (from AI-generated project description)

• Previous insights about voice preservation and practical workflows (from project memory)

• My preference for actionable techniques over theory (from analysis of my writing samples)

The AI references insights from weeks ago, builds on previous discoveries, and suggests approaches that align with everything I've established. Instead of 200 isolated conversations, I get one deepening creative partnership where insights compound instead of disappearing.

Start This Week

Which of your best AI insights have you lost in chat chaos? The three-layer system takes 20 minutes to set up and will save you hours of hunting through old conversations.

When you want to continue working, you won't hunt through chat history or re-explain basic context. You'll open your project and dive straight in—with an AI that understands your voice, your audience, and your accumulated wisdom.