Arrive at Space Lab with a workstation that can turn an idea into a working artifact. You do not need to wait for a meeting to begin. This guide uses ChatGPT and Codex as an example: start with a small project, give the agent access to its working folder, and practice building, checking, and improving something yourself.
What “agentic control” means
An agent can take steps toward a goal using the tools you connect: edit project files, run a calculation, inspect a result, and revise its work. A conversation becomes a build process when the agent has a working environment and you can examine what it produces. You choose the goal, review the evidence, and decide what is ready to use.
We use ChatGPT and Codex here because Space Lab was developed with those tools, but if you’re comfortable with another local AI agent, you can use that instead.
1. Prepare a place to work
Use a computer you are authorized to configure, a reliable internet connection, and a browser. Create a folder named SpaceLab in your Documents folder, with a separate folder for each project. Start with first-build. Keep reference material, working files, and finished results clearly named. Back up the folder before substantial changes; ask for help setting up Git when you want a history you can return to.
2. Install and sign in
Follow the official OpenAI quickstart for the current desktop app and your operating system. Sign in with your ChatGPT account and check that the features you intend to use are available on your account. In the desktop app, use ChatGPT for questions and planning; select Codex for development work. Open your first-build folder as the project so the agent has a specific place to read and change files. The quickstart also links to the terminal and code-editor options.
For the exercise below, begin with the app and a browser. Ask the agent to identify any additional tools needed before installing them. You can add Python, design applications, or hardware tools as a project requires them.
3. Give the agent a clear working boundary
Use a practice folder with copies of any reference files. Review requests for file, command, network, or application access and grant what the task needs. Connecting a folder does not automatically connect every application on your computer; a design tool may need a supported integration or a separate setup step.
Start your project with this instruction:
Work inside this project folder. Explain unfamiliar setup steps. Before installing software, changing files outside this folder, spending money, publishing, or operating connected hardware, explain the action and ask me. Keep a record of what you change and how I can run the result.
4. Build your first working artifact
Try a small rotating-habitat calculator. It needs no hardware and gives you a visible result to check. Paste this request into Codex:
Build a local browser-based rotating-habitat calculator in this folder. Let me enter habitat radius in meters and rotation rate in revolutions per minute. Show rim speed, centripetal acceleration, and acceleration as a fraction of Earth gravity. State the equations, units, and assumptions. Use a single HTML file with no external dependencies. Include a README explaining how to open it. Check the calculations, report what you tested, and tell me how to verify the result myself.
Open the resulting file in your browser. Change both inputs and watch the results. Ask the agent to explain a calculation in plain language. For a useful check, doubling rotation rate at a fixed radius should multiply the calculated acceleration by four. Zero rotation should give zero rotational acceleration. Ask it to handle blank, negative, or invalid inputs clearly. Treat the result as a learning model, with assumptions to investigate.
5. Practice the build-and-check loop
Request one improvement: a plot, a unit conversion, or a clearer explanation. Inspect the revised artifact and repeat your checks. Close and reopen it using the README. Your first milestone is being able to describe a goal, run the result, spot a problem, and ask for a useful revision. AI assistance can help you build independently; understanding and checking the result remains part of the work.
6. Bring your work to The Bog
The Bog is our shared collection of ideas, notes, tools, code, and working artifacts. Prepare a small contribution: the project files, a screenshot, a short explanation of what works, the assumptions you made, and the next question you want to explore. Remove passwords, account details, and private files before sharing. Contact Ed Kulis for onboarding and instructions on where to contribute.
Ready to join
You are ready for a productive first session when you can open your project, ask the agent to make a change, run the result, explain one check, and find your saved files again. Bring whatever you have achieved—even an unfinished experiment with a well-described problem gives us something concrete to work on.
Visit the Space Engineering Lab · Contact Ed Kulis to onboard
Setup reference: OpenAI quickstart. Product interfaces and account availability change; use the current official instructions for your installation.
