CONNECT. GET YOUR KEY. RUN ONE COMMAND.
The network has no data center — it is machines like yours, answering strangers and getting paid for it. You keep 70% of every job, earned as $GREENISH — the network's points, always worth $1 each — withdrawable in $HELLO.
Connect your Robinhood Chain wallet — MetaMask or any EVM wallet works. This is where your earnings land.
The key is your machine's identity on the network — it ties every job it answers to your wallet.
Every step below is a command you paste into a terminal window. Open one now and keep it open the whole way:
Mac — press ⌘ + space, type terminal, press enter.
Windows — press the ⊞ Windows key, type powershell, press enter.
Ollama is the free app that runs AI models on your own machine — it does the actual thinking. Paste the line for your system and press enter:
No Homebrew on your Mac? Download the app from ollama.com/download instead and open it once.
Node.js runs the small worker program that connects your machine to the network. Paste the line for your system and press enter:
On Windows: close PowerShell and open a new one after this step, so the new commands are found. No Homebrew on your Mac? Get the LTS installer from nodejs.org and click through it.
Paste this and press enter — it downloads a ~2 GB model, takes a few minutes. Same command on Mac and Windows:
Paste this in the same terminal window and press enter — it's your personal command, with your key from step 02 already in it. Leave the window open; jobs arrive on their own.
That's it. Watch the money on your earnings page — every job pays $GREENISH (1 = $1), and once you have 1 $GREENISH you can withdraw it in $HELLO with one click. Stop the worker anytime with ctrl-c; start it again whenever you want.
Text jobs need nothing beyond the steps above. But image and video jobs pay far more — an image pays 10 credits ($1.00), a video clip 50–100 credits ($5.00–$10.00) — and you keep 70%. Skip this entirely if you just want text.
The worker looks for Stable Diffusion WebUI on your machine every time it starts. If it's running, your machine serves images automatically — nothing to configure. Already have it? Start it with --api and restart your worker. Don't? Paste the line for your system:
First run sets everything up and downloads a default model — it takes a while. When it says Running on local URL, restart your worker (step 03-D) in another terminal window and it will report image models detected. From then on: just have both running.
Video is the advanced one — there's no one-click app for it yet, the way Ollama is for text. Skip this unless you already generate video on your machine (with ComfyUI or similar).
How it works: the worker doesn't make video itself. It hands each job to a video program running on your machine, waits for the finished clip, and sends it back to the network. All your program has to do is answer one simple web request:
the worker sends: POST http://localhost:9000/generate
{ "model": "my-model", "prompt": "a neon skull", "seconds": 5 }
your program replies: { "b64": "<the finished mp4, base64-encoded>" }Then start the worker exactly like step 03-D, with two extra settings in front: VIDEO_API_URL — the address your video program listens on — and VIDEO_MODELS — the name of your model (you pick it; it's the name buyers will see):
Before you run it, swap 9000 for the port your program actually uses, and my-model for your model's name. If it worked, the worker prints video models detected when it starts — from then on video jobs arrive on their own, at 50–100 credits each.
Anything that runs Ollama. A MacBook serves small models fine; a real GPU serves bigger models and gets more jobs — and can serve images and video too (see 04 above), which pay far more per job.
Electricity, and only while answering. The worker is idle silence until a job arrives — it runs quietly in the background while you do something else.
The worker only talks to Ollama on your own machine and streams text out. It runs no foreign code, opens no ports, and touches nothing else.