I used to think the people getting magic out of Claude were just better at prompting. Turns out that’s not it at all. I came across a post from an AI professional who laid out a one-time setup that makes Claude stop feeling like a stranger and start working like a real teammate. The whole idea hit me hard: the slowest way to use Claude is to keep re-prompting it from scratch every single time.
What I love about the creator’s approach is how practical it is. No fluff, no theory. Just a folder, three files, and a handful of habits. I’m going to walk you through it the way the original poster framed it, step by step, with the reasoning behind each move so you actually understand why it works.
Step 1: Switch from Chat to Cowork
The expert’s first point is the one most people miss. Regular Chat forgets you. It starts from zero on every new conversation, so you keep re-explaining who you are and what you want.
Cowork is different. It reads your files before it answers.
- Download the Claude app and open Cowork
- Let it pull context from your files instead of your memory
- Stop repeating yourself in every session
Why this matters: once Claude can read your context, the quality jump is immediate. You’re not training it from scratch anymore.
Step 2: Build the Cowork folder
This is the part I found genuinely clever. The author says a folder beats a clever prompt, and after seeing the structure, I agree.
- Make one folder called “Claude Cowork”
- Add three subfolders: about-me, outputs, and templates
- Point Cowork to that folder
That folder becomes Claude’s brain. It’s where your identity, your finished work, and your reusable formats all live in one place. The original poster’s framing stuck with me: you’re not writing a better prompt, you’re giving Claude a memory.
Step 3: Fix the settings
Here’s a detail this savvy professional flagged that explains so much. The default settings are exactly why your answers feel generic.
- Select Opus 4.8 for your hard tasks
- Turn on Thinking and set Effort to High
- Block 20 minutes and do this once
The reasoning is simple. The lighter defaults are tuned for speed, not depth. When you flip on high-effort thinking for the heavy work, Claude reasons through the problem instead of rushing to a surface answer.
Step 4: Create three sharp files
The mind behind this setup makes a great point: three sharp files beat fifty messy ones. Quality over volume.
Here’s the process the author recommends:
- Open a fresh session and ask Claude to interview you
- Use that interview to build three files: about-me, anti-ai-writing-style, and my-company
- Trim each file to under 2,000 tokens
- If a line doesn’t change how Claude responds, cut it
The rule I’m stealing from this contributor: don’t write these from scratch. Let Claude interview you and draft them, then you trim. It’s faster and the files end up sharper.
Step 5: Upgrade your habits
This section is where the original poster gets real. The setup is useless if you keep your old habits. The tools change nothing if your behavior stays the same.
- Talk, don’t type. The creator suggests installing Wispr Flow, which is free, and dictating instead of typing. You think out loud, you give more context
- End prompts with “ask me questions.” This forces Claude to fill gaps before guessing
- Call out mistakes. Don’t quietly accept a wrong answer. Correct it so the next one is better
- Start a new session every 20 messages to keep context clean and focused
That “ask me questions” tip is the one I’d start with today. It’s one line, and it flips Claude from guessing to clarifying.
Step 6: Turn your best work into a system
The final step is the one most people never reach. As this industry pro puts it, most folks quit before they build the system.
- Save your best output as a template
- Reuse that template instead of retyping the request
- Keep feeding good work back into your templates folder
Over time, your outputs folder and your templates folder compound. Every great result becomes a starting point for the next one. That’s the whole payoff: one folder, three files, and Claude that works like you, on repeat.
My honest take
What makes this work isn’t any single trick. It’s the shift in mindset the author is really selling. You stop treating Claude like a vending machine you feed prompts into, and you start treating it like a coworker who already knows your context, your voice, and your standards.
I was genuinely impressed by how low the barrier is. Twenty minutes of setup, a folder you’ll never delete, and three short files. That’s it. The rest is just better habits you build over a week.
If you want the exact file structure and the small details I couldn’t fit here, go read the full LinkedIn post from the original creator. It’s worth the few minutes.