THOX Community Mesh: Unlinkable Subscriber Tokens, Digest-Audited Stage Execution and a Witness-Cosigned Settlement Ledger for Member-Contributed Inference
Consolidated v1.1.0 edition. Most people who want to run a capable language model can afford neither the hardware nor a metered cloud account, while their neighbours' computers sit idle. This paper specifies and tests a community mesh for THOX MeshStack: members add their own devices to a shared pool and reach models through it, without the pool learning who is asking. The design combines blind-signed subscriber tokens (RFC 9474) and sealed envelopes through a relay (RFC 9180) for requester unlinkability; signed per-hop receipts with sampled replay audits for integrity; contribution credits that raise a member's token budget and tier, with no currency; and a witness-cosigned Merkle settlement ledger that settles credits and penalties after the request, off the token path. The paper states its limit plainly: a contributor that executes a stage can read its activations, so identity privacy, not content privacy, is what the pool provides, and requests that need the latter are refused rather than routed. Evidence: a real 135M-parameter model split into three stages ran token-identical to the whole model; a SHA-256 digest audit exposed 102 of 102 cheated requests; a virtual lab of 15 processes ran 12 scenarios with every lab requirement token-identical to the whole model. This edition is consolidated with the companion Community Compute paper and lab: it adopts the plan-wide admission rule, the execution contexts and the replica-first default, reproduces the companion lab's 62 tests and 150-trial campaign on a second host, and measures the replica-first claim in its own lab (whole replica 124.4 ms against two-stage pipeline 134.5 ms per remote round trip, 30 trials each, all token-exact). Supersedes v1.0.0 (DOI 10.5281/zenodo.24115596). Preprint, not peer reviewed.
Authors
- Craig Ross (ORCID: https://orcid.org/0009-0005-9872-3673)
- Tommy Xaypanya (ORCID: https://orcid.org/0009-0000-4425-9565)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-10
- DOI
- https://doi.org/10.5281/zenodo.24115595
- Primary Topic
- Privacy-Preserving Technologies in Data
- Type
- preprint