AI product
Open Compute is a compute layer from Factory for running AI agents on customer-chosen infrastructure inside the customer's own VPC. It abstracts the compute lifecycle, including creating, pausing, resuming, snapshotting, and forking environments, so agents do not depend on a fixed compute provider.
The design uses a control plane, compute gateway, relay, and pluggable drivers for environments such as EC2, Kubernetes, and sandbox providers. It also supports customer-managed model keys and is planned for open-source release.
1 use taken from transcripts — each links to the moment in the video.
A to-be-open-sourced compute layer from Factory that abstracts the compute lifecycle (create, pause, resume, snapshot, fork) for any agent, letting agents run on customer-chosen compute (EC2, Kubernetes, sandbox providers) inside the customer's own VPC via a compute gateway, relay, and pluggable drivers.
1 in the library.