A Rapper Tweeted About GitHub, So I Built Him a Learn-to-Code Repo from My Phone

Meek Mill tweeted about GitHub. It was the middle of the night. I was in bed, half asleep, scrolling on my phone.

So I messaged my AI agent from my phone, told it to build a learn-to-code repo for Meek Mill styled like album tracks, and it created the repo on GitHub and assigned the task to Copilot coding agent. Copilot wrote the content, structured the tracks, and opened a PR. I reviewed and merged from my phone.

I never opened a terminal. Never opened an IDE.

The repo: dreamchasers. Six tracks plus a bonus, album format, each one teaching a real GitHub skill. Track 1 to first repo in 10 minutes. Here's the tweet.

A rapper with 10 million followers just put "GitHub" in front of people who have never heard of a pull request. Some kid scrolling past that tweet is going to think "wait, I can do that?" and the answer is yes.

You don't need a CS degree. Don't need a bootcamp. Don't even need to leave your browser. Just a GitHub account and curiosity.

What's New in Copilot CLI

I wrote the official getting started guide back in October. A lot has changed since then.

It's Free

Copilot CLI is included as a core feature of all GitHub Copilot plans: Free, Pro, Pro+, Business, and Enterprise. You get a free GitHub account, you get Copilot CLI. Each interaction uses premium requests from your plan's allowance (50/month on Free).

Three Modes: Ask, Plan, Autopilot

Press Shift+Tab to cycle between modes:

Ask/Execute mode is the default. You prompt, Copilot acts, you approve each step.

Plan mode analyzes your request, asks clarifying questions, and builds a structured implementation plan before writing any code. Catches misunderstandings before they become bad commits.

Autopilot mode carries the task forward without step-by-step approval. You outline the work, then let it run.

Model Agnostic

Use /model to switch between Claude, GPT, Gemini, and more. Switch mid-session. Run /fleet to execute across multiple models in parallel.

Interactive and Programmatic

Two interfaces:

Interactive: Run copilot and have a conversation. Use /plan to outline work, /fleet to parallelize, /delegate to hand off, /chronicle to review your session history.

Programmatic: Pass a prompt directly with copilot -p "your task here". It runs, completes the task, exits. Good for scripts and automation.

Native GitHub Integration via MCP

Copilot CLI works directly with your issues and pull requests through GitHub's native MCP server. Search issues, analyze labels, create branches, open PRs.

Custom Agents and Skills

Use AGENTS.md and Agent Skills to define custom instructions and tool access, so your agent works the same way every time.

Copilot Coding Agent

The coding agent is what I actually used to build the dreamchasers repo from my phone. It works autonomously in a GitHub Actions environment: you assign it an issue or describe a task, it explores your repo, writes the code, runs your tests and linters, and opens a pull request. You review it like any other PR.

Copilot coding agent is available with GitHub Copilot Pro, Pro+, Business, and Enterprise plans.

The Repo

dreamchasers. Six tracks plus a bonus, album format — from making your first repo to contributing to open source. Everything runs in the browser.

Want to Try Copilot CLI?

This is the tool that built the repo. It turns plain English into code, right from your terminal.

# Install Copilot CLI
npm install -g @github/copilot

# Launch it
copilot

Included in all Copilot plans, including Free. All you need is a GitHub account.

If you want to go deeper, here's the official guide I wrote for the GitHub Blog and the full documentation.

The Point

Meek Mill getting into GitHub matters because someone who grew up on Dreams and Nightmares might see this and realize the tools are free, the community is open, and nobody is checking credentials at the door.

Dream chasers ship code. 🎤