Where it runs
The model trains and runs on your office Mac Studio. Your partners’ requests reach it encrypted, and nothing is kept along the way.
Some of what your team knows is worth sharing with the companies you work with. Teach a model the job on your own hardware, then share it with a partner by sending a link. Their Agents use it like one of their own, it keeps running on your computer, and you can stop sharing whenever you like.
“I trained the matcher on four hundred of our old orders. Now our partners’ Agents use it too.”
Three retail partners now send orders already matched to your catalog, and the model doing the matching never left your office Mac.
“Our retailers keep ordering by their own product names. Can we teach a model to match them to our catalog?”
Turns 400 past orders into examples, sets 50 aside for testing, and starts training on the Mac Studio.
“The new model matches 96% of the test orders, up from 71% before training. Put it to work?”
Clicks Share on the model and sends the link to Harbor Wines. Their Owner accepts it the same morning.
Harbor’s Agent sends over this week’s order with every line already matched to your catalog.
The model trains and runs on your office Mac Studio. Your partners’ requests reach it encrypted, and nothing is kept along the way.
Otis prepares the examples and runs the training. Your partners’ Agents use the finished model in their own work.
Each share link works once. Sharing is free, both sides see how much the model is used, and either side can end it at any time.
Your training data stays on your hardware, and your partners get the model’s help without ever getting the data.
Turn on “Train models here” in Geyser Host on an office Mac.
Explain the job to an Agent and approve the model it trains.
Click Share on the model and send the link to your partner.
Ada answers customers over email and Slack at any hour, in your voice, and passes you only the messages that need you.
Read the storyAgents answer “where’s my package?” and keep an eye on orders, so they notice when a best-seller is about to run out.
Read the storyRowan joins your calls, keeps a transcript, and sends the follow-up. Weeks later, Rowan can still answer “why did we decide that?”
Read the storySet up a workspace, give an Agent something useful to do, and see how far it gets.