How Lineage works

Compute becomes useful
when results can be trusted.

Lineage is a browser-first GPU execution design. It separates local computation, result verification, and public settlement so each layer can be measured independently.

01 / Inspect

Know the available path.

The client requests a high-performance WebGPU adapter and reads only browser-exposed capability data. NVIDIA, AMD, Intel, and Apple devices share the same run interface. WebGL2 provides a limited fallback where WebGPU is unavailable.

02 / Execute

Run with explicit consent.

The operator chooses a 15, 30, or 60 minute session and approves sustained GPU load. The run can continue behind another browser tab, but the Mine page must remain open. Closing or navigating away interrupts it.

03 / Verify

Do not trust a self-reported score.

A production coordinator must issue signed, short-lived leases with committed inputs and hidden challenges. Deterministic results can be replayed on verifier-controlled hardware before any reward is accepted.

04 / Settle

Commit covered claims.

Accepted entries can be grouped into epochs. A Solana claim program can hold vault assets, publish a Merkle root, prevent duplicate claims, and transfer only liabilities already covered by the treasury.

Hardware paths

One network interface.
Different silicon lineages.

NVIDIADiscrete GPU AMD

AMDDiscrete GPU Intel

IntelIntegrated + discrete Apple

AppleUnified silicon

Current boundary

The instrument is real.
The reward network is not live.

The current build performs actual local GPU work. It does not yet receive signed jobs, submit results to a verifier, acquire assets, or create token claims. Those systems require backend services, funded vaults, on-chain programs, security review, and operational controls.

Enter Lineage ->Read blueprint ->

Compute becomes usefulwhen results can be trusted.