Enroll and authorize
A user-installed Node identifies the device and lets participants set CPU/GPU and scheduling limits.
03 / Distributed compute
Your computer could help Eureka learn.
The Eureka Compute Network is being planned as a network of computers running genuine, verifiable AI workloads with resource limits, safeguards and honest contribution metrics. There is no public Node release yet.
GPU and CPUVerified resultsFuture rewards
Being connected is not enough. The work must matter.
The intention is not to reward energy consumption or claimed computing power alone. We aim to account for useful AI development results under quality and verification standards.
Secure task assignment
A coordinator distributes approved, limited workloads compatible with the participant’s hardware.
Execute genuine work
Inference, benchmarking, curated preprocessing, evaluation and targeted training where technically feasible.
Independent verification
Results return to the network core and are tested before acceptance or use.
Contribution credits
Approved work enters the contribution ledger and may qualify for rewards under future published rules.
Transparency dashboard
Device status, tasks, verification, availability and measurable Eureka progress.
NODE CONTROL / CONCEPT
A dashboard built around real work.
We envision configurable CPU/GPU, Node status, task history, accepted or rejected work and contribution metrics. Figures will only be displayed when backed by verified data.
Rewards should depend on verified work and an emission policy defined before launch. There is currently no guaranteed earning rate.
How rewards may workCPU / CONFIGURABLEGPU / CONFIGURABLE
MOCKUP · NO LIVE DATA SHOWN
Shared compute. Control preserved.
Distributed computing requires careful safeguards before public launch. Security must be part of the design, not an afterthought.
Workload isolation
No arbitrary third-party code execution; controlled, signed and revocable workloads.
Protected data
Private memory and personal data must not be sent to untrusted external devices.
Anti-fraud
Result verification, per-device limits and mechanisms against fake or reused work.
Controlled power use
Configurable power and hardware use limits, allowing users to stop at any time.
Model integrity
Model and central memory updates go through evaluation; an external Node cannot decide what is true.
Auditability
Work and verification records sufficient to explain credits and results without revealing user secrets.
Practical information
Before installing a Node
Can I use just my GPU or CPU?
We intend to let users choose resources and separate limits. Compatibility depends on the final release.
Will I automatically earn tokens for being connected?
No. The proposal ties credits or future rewards to genuine accepted tasks under published rules.
When can I download the program?
There is no public release, confirmed launch date or final hardware requirements yet.
Can a Node change Eureka’s memory?
Not directly. The intended architecture keeps official models and memory under central validation.