Tutorials
Most AI tutorials spend you before they teach you. Install this, copy that, paste in a key, wire up a service, and somewhere past the setup there is supposed to be a result. Plenty of people never get to see it.
Every tutorial here is a flow that already ships inside the app. Nothing to install, nothing to paste: open the flow, run it once, and read what it did.
The written walkthroughs are still being recorded. The flows below are live today.
01
Write a paper with the compiler running beside you.
Papers puts the conversation on the left and the document on the right: source, compile log and rendered output in a sidecar that stays open while you work. You ask for a section, the flow writes to the file, the compile result comes back in the same view. Writes that persist stop and ask first, so the document never changes behind your back.
- Ask for a section, get a file.
- Watch the compile as it runs.
- Approve the writes that stick.
02
Cut a reel from photos, narration and music.
Reel Machine takes the images you point at, a voice track and a piece of music, and returns a cut. Around it sit the smaller media flows you will reach for on their own: generate an image from a prompt, render frames on the machine, apply a transition between two clips, animate a still into a short video. Media work is where the local lane earns its place. The machine renders, and nothing has to leave it.
- Point at the folder, not at a file list.
- Render on the machine when local is enough.
- Keep the output in the Space that made it.
03
Pull the outside world into a Space.
Scrape a page into the Space it belongs to. Download an authenticated board into the local sandbox. Search your own mailbox and get patterns back instead of a wall of threads. Each of these lands as files inside a Space, where every other flow can already reach them.
- The web arrives as files, not as a preview.
- Credentials stay on the machine.
- What arrives is searchable immediately.
04
Let the builder build the flow.
Build designs, assembles, tests and publishes new flows through conversation. You describe the outcome; it composes atoms that already exist in the registry, runs the result, and publishes it once it works. A published flow is also a tool, so the Sim can call it the next time the same job shows up.
- Start from the outcome, not the graph.
- Reuse the atoms that already exist.
- Publish only what actually ran.
The fastest way to understand gOS is to run one flow end to end and read the trace it leaves behind.