Hi everyone,
We’re requesting $45,000 from Zcash Community Grants to build open infrastructure for inference on independently operated machines, with usage metered and settled in shielded ZEC.
Project: OpenWeights Terminal
Category: Non-Wallet Applications
Owners: @tarzelf and @zacko2o
Application: Full proposal on GitHub
Your model. Your machine. Your rules.
As AI becomes part of everyday work, control over the models we depend on matters. People should be able to choose a model, run it on hardware they control, inspect its behavior, and keep using it when a provider changes its policies or shuts down an endpoint.
That is the conviction behind our manifesto. Open weights make that independence possible. Making it practical also takes usable software, credible performance data, available machines, and a way to pay the people operating them.
OW Terminal connects those pieces. Today, the product brings together open-weight models, hardware comparisons, and source-linked throughput reports for specific model and hardware configurations. The catalog distinguishes reported results from reproduced measurements, helping people evaluate what they can actually run.
We are extending that foundation into a market for independent compute: use your own machine, or choose a host based on the model, price, performance, and availability it offers. This includes uncensored and unfiltered models, with users choosing the behavior they want and hosts deciding what they run.
Zcash has a concrete role in that system. Compute providers need to be paid, and buying inference should not require publishing a payment history on a transparent blockchain. Our proposed ZEC integration would connect metered inference to shielded payments, creating a practical reason to use Zcash for everyday work.
The grant would fund:
- Open payment specifications: payment memos, prepaid inference packs, usage accounting, and refunds.
- A local host client and reproducible runbook: a host control interface bound to the operator’s own machine.
- An open model × hardware dataset: throughput results with sources and enough configuration detail to evaluate and reproduce them.
- A verifiable shielded ZEC payment path: public implementation and acceptance tests demonstrating the agreed settlement flow.
The proposed budget allocates $25,000 to development, $10,000 to hardware/software, and $10,000 to early user acquisition. Funding is structured as $25,000 in startup funding and two $10,000 milestones.
The product is live and the network is early. At the time of this post’s preparation, the pricing page (currently just for demo purpose) listed one connected host offering inference at $1 per million tokens, marked as not yet endpoint-tested. The host revenue share is 80%. Host availability and prices can change; the grant would support completing and validating the host workflow and shielded payment integration.
We also want to be precise about privacy. Running inference on your own hardware gives you control over the machine processing your prompts. Using a remote host introduces trust in that operator. Shielded payments protect payment information; they do not, by themselves, hide prompts from the machine running the model. We are not making TEE or shielded-prompt claims.
This application covers public specifications, software, data, and payment infrastructure. It does not propose a new tradable token, ZSA, Hyperliquid integration, or changes to the Zcash protocol. Season points are a usage ledger and are outside the grant.
Our goal is to make open models useful on more people’s machines—and make shielded ZEC a practical way to pay those people for the work their machines do.
We welcome feedback from the Zcash community, especially on payment design, host onboarding, and what reviewers would want to see in a convincing settlement demonstration.
Tagging @ZcashGrants.
— Team OW Terminal (team@owterminal.com)