The golden setup: get running on AI in an afternoon

Most founders meet AI through a chat box in a browser tab. You ask, it answers, you copy the answer somewhere, the thread scrolls away, and tomorrow you start over. That’s using AI like a search engine — and it’s why it never quite compounds.

The founders pulling real leverage out of it work differently. Their AI lives on their machine, next to their actual files, and it remembers. The setup below is the on-ramp: four free (or nearly free) tools and a handful of habits. No coding. Give it an afternoon.

What you’ll install

• VS Code — a free editor from Microsoft. Think of it as the room your work happens in: files on the left, a terminal at the bottom.
• Git — the thing that saves versions of your work and syncs them to the cloud. You won’t operate it by hand; your AI will.
• Claude Code — Anthropic’s AI that works in the terminal, with permission to read and write your actual files. This is the engine.
• Obsidian — a clean reader for your notes (plain text files). It turns the folder your AI is working in into something pleasant to read.

Install them in that order. Each has a one-click installer; let the AI walk you through any step that snags.

The workflow

Once they’re installed, this is the whole loop. You’ll do it from memory within a day.

1. Open VS Code.
2. Open your project folder — keep all of them inside one dev folder on your machine.
3. Open the terminal: Ctrl + ` (the key above Tab).
4. Type claude and hit enter.
5. Tell it what you’re doing — see the two openers below.

Starting something new

Paste this on the first message of a brand-new project:

Creating a new project. Create a private GitHub repo for me. Commit and push automatically without me having to tell you. Create a memory/context system so that when we clear context we can pick up in a fresh chat on the same page. Follow best practices. FYI I am non-technical, so explain things clearly. This project is to [describe what you're building].

That one prompt sets up version control, a private backup, and a memory system — the scaffolding that lets you stop and restart without losing the plot.

Picking up an existing project

Just open the folder, run claude, and say “load up.” It reads its own memory and arrives on the same page you left.

Wrapping up a session

When you’re done for now, don’t just close the window. Type:

Let’s wrap up here. Commit, push, and prep for context clear.

It saves your work, backs it up, and writes down where you got to — so the next session starts clean instead of cold.

Stay under the line

AI gets worse as a single conversation gets long — it’s holding too much at once. Two commands keep you sharp:

• /context — shows how much the current chat is holding. Scroll to the colourful grid; the number to its left is your tokens used. Stay under ~200k. Past that, quality drops and usage burns fast — wrap up and start fresh.
• /usage — shows how much of your daily and weekly allowance you’ve spent.

Three habits that compound

• Feed it files, not paste. Instead of dumping context into the chat, drop it in the project folder and say “read this, pull out what matters.” The chat stays light; the knowledge stays put.
• Ask for files, not answers. Rather than have it write a long reply you’ll lose, say “write that to a markdown file.” Markdown is just text — and now it’s saved, searchable, reusable.
• Read in Obsidian. Open your dev folder as a vault and your notes render clean. Now you can actually see what your AI is building with you.

That’s the setup

Four tools, one loop, two commands, three habits. It looks like a developer’s desk because it is one — and that’s the point. This is the foundation everything else at 1000xfounder sits on: your business as one place your AI can read, with you in command of it.

This is also exactly how 1000xfounder itself gets built — in public, on this same setup.