1 / 17
Every AI has a memory limit
Context Window
New data comes in, old data gets pushed out
The File Limit Wall
50 transcripts
LIMIT: 10-15
Accepted
Analysis Falls Apart
Quality
Data Processed
The Solution: External Notes
📄
Write → Clear → Read back → Continue
Files that persist when AI memory resets
The Four Components
📁
Source Data
Transcripts or emails in a folder
📋
Context File
Goal reminder read after each reset
Todos File
Tracks what's done and what's left
💡
Insights File
Final output with discoveries
The Cycle
Process Files
Work
Update Notes
Save
Memory Clears
Reset
Read Notes
Resume
20-40 minutes of continuous work, quality stays consistent
Setting Up Claude Code
Use Case 01
Customer Language
Stop guessing what words to use
What Claude Extracts
Extract
💢
Exact phrases describing problems
Questions they ask
😟
Concerns & hesitations
🔍 Filter: frustration, stress, fear, confusion
I want you to analyze all the meeting transcripts in this folder to find patterns in how clients describe their problems, what questions they ask, and what concerns they raise. If it does not cause frustration, stress, fear, or confusion, it does not count. Before you start: 1. Create a context markdown file that contains the goal of this analysis: extracting customer pain points in their own words for future content creation 2. Create a todos markdown file to track which files you've analyzed and what you've found 3. Create an insights markdown file that you iteratively update after processing each transcript As you work: - Iteratively update the insights file after processing each transcript - Check off each transcript in the todos as you complete them and make sure it's updated before your memory gets compacted - After any memory compaction, read context and todos files before continuing For each transcript, extract: - Exact phrases used to describe problems or pain points - Questions asked - Concerns or hesitations mentioned Work through all files until complete.
Use Case 02
FAQ Building
Every question represents friction
The Friction Multiplier
?
1 question asked
=
?
?
?
?
?
?
?
?
?
?
10+ who didn't ask
Some walked away instead
What Claude Extracts
Extract
Every question asked
🏷️
Topic context
💬
How you answered
🔄
Follow-up questions
🔍 Filter: confusion, uncertainty, gaps in understanding
I want you to analyze all meeting transcripts in this folder to extract questions clients and prospects have asked. Focus on questions that indicate confusion, uncertainty, or a gap in their understanding. Before you start: 1. Create a context markdown file that contains the goal: building a comprehensive FAQ from real customer questions 2. Create a todos markdown file to track which files you've analyzed 3. Create an insights markdown file that you iteratively update after processing each transcript As you work: - Iteratively update the insights file after processing each transcript - Check off each transcript in the todos as you complete them and make sure it's updated before your memory gets compacted - After any memory compaction, read context and todos files before continuing For each transcript, extract: - Every question asked by the client or prospect - The context of the question (what topic it related to) - How the question was answered (if an answer was given) - Any questions that required follow-up or clarification Work through all files until complete.
The Common Structure
Setup
Create context.md
Create todos.md
Create insights.md
Work Loop
Update insights + todos as AI works
Check off todos before compaction
Read context + todos after compaction
Customize
Your goal
What to extract
What matters most
Same skeleton, different extraction targets
Same Approach, Endless Uses
⚠️
Churn Risk
Spot complaints before they leave
💡
Feature Ideas
What they keep asking for
🔥
Lost Leads
Follow-ups you forgot
Any large dataset • Transcripts • Emails • Tickets • Documents
The Takeaway
📋 Context
+
✅ Todos
+
💡 Insights
=
Unlimited Memory
01
Write notes before memory clears
02
Read notes after memory resets
03
Quality stays consistent