Validate

Validators score every agency (miner) by the live audience of its tutors’ lectures. They check Telureo’s numbers against YouTube, and set weights on the subnet.

Join

First, point the command line at Telureo, and keep your computer’s clock synchronized. Signed requests are refused when it’s more than a minute off.

export EDU_API_URL=https://web-production-eecc0.up.railway.app

1. Contact the subnet owner. They add your hotkey to the validator list, and invite your Google account to Telureo’s channels as “Viewer (limited)” in YouTube Studio.

2. Connect YouTube. It’s optional, and recommended for spot-checks. In your own Google Cloud project, enable the YouTube Analytics API, and create an OAuth client of type “Desktop app”. Then sign in once with the invited account:

   edu validator connect-youtube --client-file client_secret.json

Run

Every 30 minutes, the validator downloads the dataset, scores it with the public scoring code, sets weights, and spot-checks a few lectures against YouTube:

edu validator run --netuid <netuid>                       # --network defaults to finney
edu validator run --netuid <netuid> --dry-run --rounds 1  # print the weights without setting them

Other commands, for a closer look:

edu validator fetch --out data.json                  # save the dataset
edu validator score --file data.json --netuid <netuid>  # score a saved dataset
edu validator check --lecture <id>                   # compare one lecture with YouTube

Spot-checks, in run and in check, need connect-youtube first. A difference is a red flag only if it’s still there after Telureo’s next refresh.

How weights are set

• A lecture counts from 48 hours to 120 hours after it actually ends, a 72-hour scoring window.
• Its score is its latest live watch-minutes. A value fetched 60 minutes ago or more doesn’t count until a new one arrives.
• An agency’s score is the sum of its lectures’ scores. Its weight is its score divided by the total of all agencies’ scores, in exact arithmetic.
• Burn. When every score is zero, or a valid dataset isn’t available, the validator puts all weight on the subnet owner’s UID.

Validators work from the same Telureo data, so their weights agree, and any validator can confirm another’s finding against YouTube.