Platform Lab
Learn to turn one machine into a platform you can reach from anywhere, publish to the internet safely, restore when it breaks, and that tells you first when something goes wrong. Setup and checks are handed to an AI coding agent; you learn what to hand over and what to check yourself. The danny22.com you are reading is the result.
- Status
- Every unit has a first draft (2026-10-27). The prompts in each unit were tested; the material is now being revised
- Size
- 11 parts, 39 units, each about an hour including hands-on work
- You need
- A DGX Spark class machine, a monitor, keyboard, mouse and KVM switch, a laptop and a domain, plus accounts for a DNS and tunnel service, a VPN service and a code hosting service, and an AI coding agent subscription
- Approach
- No ready-made project to download. Following the units, learners build their platform step by step by giving prompts to the AI coding agent on their own machine, then run it the same way
- Access
- platform-lab.danny22.com, open to signed-in readers. Get in touch to request access
- Language
- The course material is written in Korean
What you learn
- Reaching the machine from anywhere: get in from outside the home over a VPN, and prepare a way back in for when the VPN fails.
- Publishing safely: a tunnel that needs no open ports, a reverse proxy that sends each address to the right service, domains, DNS and https.
- Letting in only people you know: manage people's logins (with two-factor authentication) and programs' API keys in one place, and send invitation and password reset mail.
- Restoring when it breaks: keep encrypted backups off the machine, test regularly that they really restore, and pin versions to upgrade one at a time.
- Knowing first: regression tests, failure alerts, a heartbeat that warns you from outside even when the machine is off, and monitoring that keeps checking the services.
- Handing operations to an AI agent: grow the agent's rules and operating knowledge together, decide how far to delegate and where a person approves, and keep secrets out of conversations and screens.
What a unit contains
- Theory: why it is built this way, including real incidents and how they were caught.
- Procedure: each step gives a prompt to paste as is, what the agent does, and what a person does (accounts, approvals, checks). Parts 1 and 2, setting up the machine and reaching it from the laptop, are done by hand.
- Checks: a unit is done when its check commands and regression tests pass.
- Lines for your rules: what each unit adds to the agent's rules file, so your own rules grow with the course.
- Reference answers: configuration extracted from this running platform with only the values replaced, for comparing when you get stuck.
Parts
| Part | What you learn | Units |
|---|---|---|
| 0 Destination | What you will have at the end, what you need when, and what it costs | 3 |
| 1 Bringing the machine home | What the machine is good at, a desk where the monitor and keyboard are shared with the laptop, first boot, and joining the machine and laptop over a VPN so they find each other by name | 3 |
| 2 Getting in from the laptop | SSH server and keys, and a remote development setup where the screen stays on the laptop while files and terminal live on the machine | 3 |
| 3 Bringing in the agent | An AI coding agent on the machine and its first rules, version control and code hosting, repository design, container compose and a command runner, environment variables and secrets | 6 |
| 4 Getting in from anywhere | Reaching the same setup from outside, having the agent check the VPN, an out-of-band access path for when the VPN fails, and drawing the connections | 2 |
| 5 First service | Running a service inside the machine before going public, ports and binding, memory and disk checks, and a regression test harness set up before going public | 3 |
| 6 Going public | Domain and DNS, a tunnel with no open ports, a reverse proxy that receives every address, and an https policy | 4 |
| 7 Only people you know | An identity service for people's logins and programs' API keys, an API gateway and a first API service, and invitation and password reset through a mail sending service | 6 |
| 8 Keeping it safe | Off-machine backups and restore tests, pinned versions and a safe upgrade order | 3 |
| 9 Knowing first | Failure alerts, heartbeats watched from outside the machine, and monitoring of the services | 2 |
| 10 Running it for the long haul | Operating knowledge the agent carries across sessions, widening delegation, adding a service and migrating an existing system, and a look back | 4 |