---
title: "I looked at 30 days of my AI conversations and found something surprising"
description: "5 atomic claims about a detailed analysis of 30 days of chatgpt and claude conversations reveals 10 repeating prompt patterns that demonstrate systematic ai use. the author shares specific prompt structures for tasks like email triage, presentation assembly, and workflow documentation, showing how to treat ai as infrastructure rather than a casual tool."
url: "https://kbanc.com/claims-library/30-days-ai-conversations-surprising-patterns"
date: "2025-10-22"
topics: ["strategy", "implementation", "tools"]
generated: "2026-02-19"
---

# I looked at 30 days of my AI conversations and found something surprising

By Kamil Banc | October 22, 2025

## Claims

1. **10 patterns emerged from analysis**: The author identified 10 distinct repeating patterns in 30 days of AI conversation history across ChatGPT and Claude
2. **Email triage identifies priority actions**: Email triage prompts filter inbox to identify what needs response today, who's waited 48+ hours
3. **Prompt merging creates reusable infrastructure**: Prompt optimization merges multiple templates into single reusable tools under 200 words for varied cases
4. **Custom skills automate recurring tasks**: Custom skills enable repeatable workflows like morning briefings analyzing 7 days of Gmail on command
5. **Infrastructure mindset drives AI effectiveness**: Effective AI prompts specify context, constraints, output format, and exclusions as systematic infrastructure

## Evidence

### Quote
> "None of these prompts ask AI to think for me. They ask AI to execute plans I've already made. Every prompt includes context, constraints, and desired output format." - Kamil Banc

### Key Statistics
- **30 days**: Period of AI conversation history analyzed to identify systematic usage patterns
- **10 prompt patterns**: Distinct categories of repeating prompt structures identified from the analysis
- **500 character limit**: Content adaptation constraint for converting long-form technical content to Substack Notes format
- **200 words total**: Maximum length requirement for merged, reusable prompt templates

## Context
The analysis methodology involved pulling 30 days of prompts across ChatGPT and Claude, then categorizing them to identify repeating patterns. Each prompt type was anonymized and simplified to show the structural approach rather than specific content. The author provides a meta-prompt that readers can use to run the same analysis on their own conversation history, identifying task types, output formats, recurring workflows, and automation opportunities. This diagnostic approach reveals how users are building systems without explicitly recognizing them as automation, allowing for optimization and template creation. The article concludes with a specific audit prompt that groups conversations by task type, frequency, and optimization potential.

## Source
- Original: [I looked at 30 days of my AI conversations and found something surprising](https://aiadopters.club/p/30-days-ai-conversations-surprising-patterns)
- Cite: kbanc.com/claims-library/30-days-ai-conversations-surprising-patterns