· 8 min read
Christian Acuna — Terminal as Your AI Workshop
Learn five beginner-safe terminal commands and create your first local Founder OS workspace for Claude Code, Finder, and Zed.
Written in by Christian Acuna Founder, AI-Native Founder
Foundations Lesson 2 of 3
This lesson is part of the Foundations sequence. Read it in order or jump back to the table of contents.
Your agent needs somewhere real to work.
A chat window can answer questions. A workspace can hold context, files, decisions, tools, and review loops. That is the difference between asking AI for help and building an AI-native operating environment.
This foundation gets your Mac ready for that environment.
You are going to install a small operator stack: a terminal, Apple’s command line tools, a package manager, Node.js, an IDE, Claude Code, and a first project folder with CLAUDE.md.
Do not worry if some of those words are new. The goal is not to become an engineer overnight. The goal is to give yourself a clean place where an agent can read, plan, edit, and help you ship.
pwd, ls, mkdir, cd, and open ..This guide is Mac-first. Windows and Linux workflows are possible, but branching too early makes the first setup harder than it needs to be.
A few rules before we start:
Here is the simple mental model:
The model is not the whole system. The system is the model plus a real workspace it can understand.
Your terminal is where you talk to your computer directly.
macOS already includes Terminal.app. You can use it. For this path, I recommend Ghostty because it is modern, fast, and visually clean enough that beginners do not feel like they are fighting the tool.
Now run your first command:
pwd
pwd means print working directory. It tells you where the terminal currently is.
You have not built anything yet. That is fine. The first win is opening the control room.
Many developer tools on macOS expect Apple’s command line tools to exist. Homebrew may prompt for them automatically, but it is better to install them directly.
In Ghostty, run:
xcode-select --install
A macOS dialog may appear. Follow the prompts.
When it finishes, verify with:
xcode-select -p
If you see a path printed back, you are good.
If the install fails or Apple says the tools are unavailable, use Apple’s developer downloads page: developer.apple.com/download/all.
Homebrew is the package manager for macOS. It lets you install and update tools without hunting for random downloads.
Open the official page: brew.sh.
At the time of writing, the Homebrew install command is:
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
Run it in Ghostty.
Homebrew may print a few commands at the end telling you to add Homebrew to your shell profile. Copy and run the exact commands Homebrew prints for your machine.
Then verify:
brew --version
brew doctor
If brew doctor prints warnings, do not panic. Warnings are common. We mostly want to confirm that brew exists and can run.
Homebrew’s installation docs are here if you need the source of truth: Homebrew Installation.
Node.js powers a huge amount of modern web and AI tooling. nvm means Node Version Manager. It lets you install and switch Node versions without breaking your system.
Open the official nvm repo: github.com/nvm-sh/nvm.
Use the latest install command from that page. After installation, close and reopen Ghostty if nvm is not found.
Then install the long-term support version of Node:
nvm install --lts
nvm use --lts
node --version
npm --version
You should see version numbers printed back for node and npm.
That means your Mac can now run the JavaScript tooling that many modern products, websites, and agents depend on.
An IDE is where you read and edit the files in your workspace.
We will start with Zed because it is fast, simple, and less visually noisy than many fully loaded development environments. VS Code is also fine, but standardizing on one editor keeps the learning path clean.
If Zed offers to install command line integration later, you can accept it, but this lesson does not depend on that.
Claude Code is the agentic coding assistant that runs from your terminal and works inside your project folder.
You will need access through a Claude subscription, Claude Console account, or supported provider. Anthropic’s docs are the source of truth:
At the time of writing, the native install command for macOS, Linux, and WSL is:
curl -fsSL https://claude.ai/install.sh | bash
After install, verify:
claude --version
Then launch Claude Code:
claude
Follow the authentication prompts in your browser or terminal.
If claude is not found, close and reopen Ghostty, then try again. If it still fails, return to Anthropic’s setup docs and follow the path update instructions for your shell.
You do not need to memorize the entire terminal. Start with five commands.
pwd — show where you arepwd
This prints the current folder.
ls — list what is herels
This shows files and folders in the current location.
mkdir — make a foldermkdir Projects
This creates a folder named Projects.
For nested folders, use -p:
mkdir -p ~/Projects/ai-native-workspace
cd — change folderscd ~/Projects
This moves the terminal into your Projects folder.
To go into your workspace:
cd ~/Projects/ai-native-workspace
open . — open the current folder in Finderopen .
The dot means “this current folder.” On macOS, open . opens the current folder in Finder.
These five commands are enough to orient yourself, create a workspace, move into it, and inspect it visually.
Now make the folder where your first AI-native work will live.
Run:
mkdir -p ~/Projects/ai-native-workspace
cd ~/Projects/ai-native-workspace
pwd
open .
You now have a real workspace directory.
Create a few folders inside it:
mkdir -p decisions projects examples
ls
You should see:
decisions
examples
projects
Now open the workspace in Zed:
zed .
If zed . does not work yet, open Zed manually and choose the ~/Projects/ai-native-workspace folder.
CLAUDE.mdInside Zed, create a file named:
CLAUDE.md
Paste this starter version:
# CLAUDE.md
## Project
This is my first AI-native founder workspace.
## Goal
Help me turn ideas into clear plans, drafts, experiments, and shipped artifacts.
## Working Style
- Plan before making changes.
- Keep steps small and reviewable.
- Explain assumptions.
- Do not invent facts, links, metrics, or sources.
## Privacy
- Do not expose API keys, passwords, private customer data, or sensitive personal information.
- Ask before using private material in public-facing work.
## Quality Bar
Output should be clear, useful, and specific enough that a founder can act on it.
This file is not magic because of the text itself. It is powerful because it creates durable context.
Instead of repeating your expectations in every chat, you put the operating rules in the workspace where the agent can read them.
Go back to Ghostty. Make sure you are inside the workspace:
cd ~/Projects/ai-native-workspace
pwd
ls
Then start Claude Code:
claude
Try this first prompt:
Read this workspace and tell me what is here. Then propose three small founder workflows I could build in this folder.
The first win is not writing code.
The first win is seeing the agent understand a real local workspace: folders, files, instructions, and a working context it can return to.
That is the beginning of AI-native work.
command not found: brewClose and reopen Ghostty. If it still fails, Homebrew likely printed shell profile commands at the end of installation. Return to brew.sh or rerun the installer and copy the profile commands it prints.
command not found: nvmClose and reopen Ghostty. If that does not work, return to the nvm install docs and follow the shell profile instructions for zsh.
command not found: claudeClose and reopen Ghostty. If it still fails, return to Claude Code setup and follow the installation/path instructions.
Good. Move on.
Read the prompt. If it is asking for access to files or folders you intentionally opened, allow it. If you are unsure, pause and check what app is asking.
You now have the basic operator stack:
~/Projects/ai-native-workspace.CLAUDE.md.That maps directly back to the flywheel:
CLAUDE.md.Setup is not busywork.
It is the moment your work moves from scattered prompts into an environment that can compound.
A folder can hold decisions. A repo can hold history. A CLAUDE.md can hold operating rules. A terminal can run checks. An IDE can show you what changed. Claude Code can work across all of it.
That is the difference between using AI and operating AI-native.
When you are ready, continue to the next foundation: terminal fluency and navigating your workspace.
Continue Foundations
· 8 min read
Learn five beginner-safe terminal commands and create your first local Founder OS workspace for Claude Code, Finder, and Zed.
· 11 min read
Scattered AI chats do not compound. Build a founder command center: one place for context, agents, review loops, decisions, and safe AI-native execution.
· 7 min read
The starting point for the Foundations path: how AI-native founders move from one-off prompting to a repeatable Context → Plan → Build → Ship operating loop.