Groundwork

$ groundwork --start_

Overview

Start as a vibe coder. Learn to think like an engineer.

You can build real software with AI today, before you know a single keyboard shortcut. That's the on-ramp. Groundwork is written for adult learners: start with a quick win, then build the hours, taste, and verification habits that move you toward "I shipped this and I can defend every line."

Quick start

Want a configured machine now? On a Mac, run the bootstrap below; on Linux, WSL, or a headless workspace, follow Platforms. Want to understand what it did? Start with The path, then use the reference pages when a tool earns its place.

Get the repo on GitHub

Same path, different environment

macOS gets the full desktop setup: Homebrew apps, browser helpers, Raycast, Anybox, BetterDisplay, and selected system defaults. Linux/WSL and headless workspaces share the terminal, Git, editor, shell, and agent-instruction layer while skipping Mac-only GUI pieces. See Platforms for the exact split.

bootstrap
$ bash -c "$(curl -fsSL https://raw.githubusercontent.com/jasondockery/groundwork/main/bootstrap-mac.sh)"

#The two modes you're moving between

Groundwork treats AI-native development as a climb, not a gatekeeping exercise. Start with momentum. Then build the judgment, habits, and verification muscles that let you keep moving fast without losing the plot.

the floor

Vibe coding

Talk to an agent, accept, run, paste the error back, and ship something that did not exist before. It raises the floor: more people can build sooner. Its ceiling is real: you may not yet understand the system you created.

the ceiling

Agentic engineering

Orchestrate agents while staying accountable for the spec, quality bar, tests, security, and architectural taste. Same speed, professional responsibility. That is the discipline this guide helps you practice.

The useful framing from Andrej Karpathy is simple: vibe coding raises the floor; agentic engineering raises the ceiling. Groundwork is for the person who wants both: the joy of building early and the chops to defend the work later. Read more in Vibe → agentic.

#The path at a glance

You do not learn these tools by reading a catalog. You learn each one when it solves a pain you can feel. The docs are now organized around that journey.

1
Ship something today
No tmux, no Vim, no theory. Open an editor, talk to an agent, build a tiny real thing.
Start with Getting started, Editors & AI, or an interest track: FPS Browser, FPS Unity, Web & sites, Apps, or Image & video.
2
Don't get lost
When vibe coding breaks and you need to navigate without guessing.
Learn paths, files, shell commands, and Git basics in Command line & Unix, Shell & CLI, and Git.
3
Earn each tool
When the agent is plausibly wrong, terminals multiply, editing feels slow, or context keeps disappearing.
Reach for tmux, Neovim, agent rules, and the deeper Git workflows only when the pain makes them useful.
4
Build the chops
When you are ready to get fast on purpose.
Use the Practice plan for short daily reps: shell movement, tmux, editor motions, Git diffs, and agent supervision.
5
Think like an engineer
When "it works" stops being enough.
Practice architecture, design patterns, specs, tests, evals, security, and taste through Development, Starting a project, and Vibe → agentic.
6
Work for real, teach others
When the loop needs to hold up for a team, a class, or future you.
Use Everyday workflow, Knowledge base, and AI agents & rules.

Follow the full path →

#Why the hours still matter

AI can draft the code. It cannot give you understanding by osmosis. The reps in this guide are how you notice when an agent edited the wrong file, introduced a brittle abstraction, skipped a security boundary, or wrote a test that proves too little.

The deal

Ship early. Then put in the reps. The tools fade into the background, your judgment moves to the front, and you build toward moving at AI speed without lowering the bar.

#Use the docs two ways

why

Why Groundwork

The case for an AI-native development foundation that teaches judgment, not just tools.

 journey

The path

A staged route from first build toward agentic engineering.

 reference

Cheat sheet

The keyboard and command shortcuts once you know what you are practicing.

 machine

Setup & chezmoi

What gets installed, how apply works, and how to inspect changes.

 ai-native

AI agents & rules

Project instructions, adapters, skills, and the shared north star.

 track

FPS Browser

Build a substantial Three.js arena shooter with AI, read the JavaScript, tune game feel, and ship a URL.

 track

FPS Unity

Build the same hard one-week shooter arc in Unity 6 with C#, prefabs, NavMesh, UI, profiling, and WebGL.

 track

Web & sites

Build a single page, style it, add JavaScript, fetch data, and deploy it.

 track

Apps

Build a small task app with a spec, database, tests, diff review, and deploy.

 track

Image & video

Build a reproducible generative-media pipeline with model APIs, seeds, manifests, evaluation, and shipped assets.

 learning

Development

Deeper routes for Apple apps, web, games, Rust, Python, ML, and LLMs.

 outside study

Resource library

Books, courses, videos, and practice options for going deeper.
Next →The path