Tired of Repeating Myself to AI Every Single Conversation

Updated January 2026 | 4 min read

You've re-explained your business 100+ times.

Your name. Your role. Your clients. Your tone. The way you structure proposals. The pricing tiers. The three clients you can't name in examples because of NDAs. The fact that you don't use Oxford commas in headlines but you do in body copy.

Every. Single. Conversation.

You're not training an assistant. You're speed-dating a stranger with amnesia.

The Problem: Every Conversation Starts From Zero

ChatGPT doesn't remember you. Neither does Claude, unless you're using Claude Code with proper context files.

Each new chat is a blank slate. No memory of yesterday's conversation. No record of the 47 times you've explained that your target audience is B2B SaaS founders, not general "business owners."

You open a new chat. You need a proposal. So you type:

"I run a video editing service for SaaS companies. We charge $3,000/month for 4 videos. Our turnaround is 5 business days. Write a proposal for a client who needs product demo videos."

Next day, different task. Same preamble:

"I run a video editing service for SaaS companies. We charge $3,000/month for 4 videos. Our turnaround is 5 business days. Write an email to a prospect who ghosted after the discovery call."

Week later:

"I run a video editing service for SaaS companies. We charge $3,000/month for 4 videos. Our turnaround is 5 business days. Create a LinkedIn post about our new testimonial from [Client Name]."

You're not working with AI. You're babysitting it.

Why This Happens

AI models are stateless. They don't retain information between sessions unless you explicitly give them persistent context.

ChatGPT's "memory" feature is unreliable. It picks up random details — your dog's name, the city you live in — but misses critical business context. It's not designed for professional use. It's designed for casual chat.

Custom instructions help, but they're limited to 1,500 characters. That's enough for surface-level details, not real operational context.

You need more than a paragraph. You need a file.

The Fix: One File That Loads Automatically

CLAUDE.md is a markdown file that contains everything about you, your work, your preferences, and your business.

When you use Claude Code (Anthropic's official CLI) with Obsidian, Claude reads this file automatically at the start of every session.

No prompts. No reminders. No re-explaining.

You write it once. Claude remembers forever.

What Goes in CLAUDE.md

Who you are. What you do. How you work.

Your company details: services, pricing, packages, constraints. Your clients: industries, personas, NDAs. Your voice: tone, style, banned phrases, preferred sentence structure. Your tools: CRM name, platforms, integrations, workflows.

Example from a real CLAUDE.md file:

Voice: No fluff. No "Let me unpack that." Use contractions. Vary sentence length. Short for impact. Don't wrap threads in conclusions.

Pricing: $3K/month for 4 videos, 5-day turnaround. $500 rush fee for 48-hour delivery. No equity deals, no trade, no payment plans under $10K total project value.

Clients: B2B SaaS only. No ecommerce, no personal brands, no local businesses. Current clients include [Company A], [Company B], [Company C] — all under NDA, never name them in examples.

One file. One afternoon to write it. Then it's done.

What Changes Immediately

You open Claude Code. You type: "Write a proposal for the Acme Corp opportunity."

No preamble. No context dump. Claude already knows your pricing, your tone, your deliverables, your structure.

The output matches your voice. It includes the right package details. It avoids the phrases you've banned. It structures the proposal the way you actually send them.

Next conversation: "Draft a follow-up email for the prospect who went dark after the discovery call."

Claude knows your email style. It knows you don't use pushy language. It knows you include a calendar link in every outreach. The draft is 90% ready.

You're not re-explaining. You're working.

How This Works Technically

Claude Code is Anthropic's official command-line interface for Claude. It runs locally on your machine.

When Claude Code starts, it checks for a CLAUDE.md file in your working directory. If it finds one, it loads the entire file as context for that session.

Obsidian is a markdown-based note-taking app. Your CLAUDE.md file lives in your Obsidian vault alongside your other notes, templates, client files, and project documentation.

Claude doesn't just read CLAUDE.md. It can read any file in your vault. Your client briefs. Your past proposals. Your content templates. Your meeting notes.

File-based context beats chat history every time.