Your machine.
Clear boundaries.
A shared grid needs clear rules. Here is what we collect, what the executor reads, and who can see a workload.
This website
Early-access sign-up is not open yet. This website does not collect or store sign-up details. The capacity estimator runs in your browser. It does not scan your machine or send the values to discomp. This site has no advertising trackers or analytics cookies. Cloudflare serves the site and processes the connection data needed to deliver pages.
Planned early-access sign-up
We plan to collect your email, name, whether you are an individual or a company or lab, and the number of machines. For each kind of machine, we plan to collect the OS, processor, logical cores, RAM, and quantity. We will use these details to verify your email, plan machine support, and send enrollment updates.
The sign-up flow, email verification, storage, and entry removal are planned. We will publish the storage and retention details before sign-up opens.
When you enroll a machine
Enrollment will require a verified email and a signed owner agreement. The agreement requires honest results and no tampering. One account is allowed per person. Enrollment and the installer are planned; an early-access entry does not enroll a machine.
What the executor reads
The executor reads system counters: CPU use, free memory and disk, memory pressure, idle time, power, and temperature. It reads a hardware summary: OS, processor model, cores, RAM, disk size, and hypervisor.
It never reads the owner’s files, apps, browsing, or keystrokes.
What the executor sends
It sends the hardware summary, the system counters above, the capacity offer, and a node ID to the control plane. The node ID is a SHA-256 hash of the OS machine ID. It never sends the raw OS machine ID.
The product design excludes hostnames from community machines. Company fleets may send them. Hostname exclusion is part of the planned enrollment release.
What a workload can see
Each task runs in its own microVM. A VM sees none of the owner’s files. It has its own scratch disk, capped at the lease size. The VM’s scratch disk is deleted when the lease ends.
The planned community network allows internet access only, with no access to the owner’s LAN or host. Company site networks have a separate design for company resources. A microVM does not grant a workload access to the owner’s desktop or apps.
Who can see workload code
A machine owner with admin rights controls the hypervisor. They can inspect or change the code in a VM. MicroVM isolation protects the owner’s files; it does not hide workload code from that machine’s owner.
The shared grid uses verified owners under an agreement. The Release 1 design keeps long-lived secrets out of owner machines and uses short-lived tokens for one job. It also checks a sample of results on a second node.
A team can choose its own machines for private code. A confidential hardware tier is a later design, not part of Release 1.
Your machine comes first
Lending follows the machine’s conditions: AC power, a usable hypervisor, temperature, and memory pressure. Idle-all mode also checks owner activity. There is no pause or resume control. The owner-first design stops an affected VM when a condition ends.
The planned installer adds a lid helper only with the owner’s consent. Uninstall will remove the service, executor, helper, data directory, and node certificate.
Contribution and rewards
The estimator shows a default CPU and RAM proposal. It does not measure your machine, free disk, or current conditions. It does not show earnings. Compute-credit rewards are planned. The rate is not set, and cash payments are not part of Release 1.