{"id":51604724,"url":"https://github.com/name970/Protocol","last_synced_at":"2026-07-31T02:00:55.843Z","repository":{"id":368812915,"uuid":"1284999986","full_name":"name970/Protocol","owner":"name970","description":"A Proof-of-Useful-Work consensus where the \"work\" is verifiable matrix multiplication: exact INT8 arithmetic for bit-reproducible results, plus a cheap probabilistic audit that certifies a product without recomputing it.","archived":false,"fork":false,"pushed_at":"2026-07-13T08:41:33.000Z","size":1163,"stargazers_count":0,"open_issues_count":5,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2026-07-13T10:20:25.068Z","etag":null,"topics":["consensus","cryptography","rust","verifiable-computation","zero-knowledge"],"latest_commit_sha":null,"homepage":"","language":"Rust","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/name970.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-30T11:42:04.000Z","updated_at":"2026-07-13T08:41:37.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/name970/Protocol","commit_stats":null,"previous_names":["name970/protocol"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/name970/Protocol","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/name970%2FProtocol","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/name970%2FProtocol/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/name970%2FProtocol/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/name970%2FProtocol/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/name970","download_url":"https://codeload.github.com/name970/Protocol/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/name970%2FProtocol/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":36098609,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-07-20T02:08:10.276Z","status":"online","status_checked_at":"2026-07-31T02:00:06.731Z","response_time":112,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["consensus","cryptography","rust","verifiable-computation","zero-knowledge"],"created_at":"2026-07-12T01:00:27.452Z","updated_at":"2026-07-31T02:00:55.808Z","avatar_url":"https://github.com/name970.png","language":"Rust","funding_links":[],"categories":["Cryptography"],"sub_categories":["Zero Knowledge Proofs"],"readme":"# Protocol\n\n\u003e A Proof-of-Useful-Work consensus where the \"work\" is verifiable matrix multiplication:\n\u003e exact INT8 arithmetic for bit-reproducible results, plus a cheap probabilistic audit that\n\u003e certifies a product without recomputing it.\n\n\u003e [!WARNING]\n\u003e **Status: research / proof-of-concept. Unaudited. Not for production use.**\n\n---\n\n## Why this is interesting\n\nQuantizing to INT8 is the canonical *lossy* step in machine learning. This project uses the\nsame INT8 path the other way around: it reconstructs a **bit-exact** floating-point matrix\nproduct from bounded integer slices. Exactness is the point — a result reproducible down to\nthe bit is the precondition for *verifying* a computation cheaply (a Freivalds-style check,\n`O(n²)`) instead of recomputing it (`O(n³)`). On top of the exact product sits a commit–reveal\nscheme, so a prover can generate its own challenges non-interactively without being able to\ngrind a passing forgery.\n\nThe decomposition technique is **not novel**: it builds on the Ozaki scheme and the line of\nwork putting it on integer / INT8 tensor cores. The contribution here is using it as the\ncompute layer of a verifiable-work consensus, with the audit and soundness analysis that requires.\n\n## What is open\n\n- Exact `INT8 → INT32` partial products with `FP64` reconstruction — implemented and covered by tests (bit-exact against a direct reference).\n- Non-interactive commit–reveal challenge derivation — designed, not yet implemented.\n- Post-quantum time-seal (delay function) — open design question.\n- Data-availability layer (FRI) and private transaction layer — specified, not implemented.\n- GPU kernels — current code is a CPU reference.\n- **Formal security proofs.** Soundness of individual pieces is being discussed in the open\n  (see [Cryptography, honestly](#cryptography-honestly)); none of it is audited.\n\n## Quick start\n\n```bash\ngit clone https://github.com/name970/Protocol\ncd Protocol\n\ncargo test                    # exactness + forgery-rejection tests\ncargo run --example demo      # bit-exact INT8 matmul vs an f64 reference (adjust to your layout)\n```\n\n## Architecture\n\nA short map from whitepaper sections to code lives in [ARCHITECTURE.md](ARCHITECTURE.md).\n\n## Contributing\n\nContributors are very welcome — especially people who want to write the code.\n\n- New here? Start with the **`good first issue`** label: self-contained tasks that touch the\n  numerics and need no protocol-wide knowledge.\n- Read [CONTRIBUTING.md](CONTRIBUTING.md) for dev setup and PR conventions.\n- Questions and design discussion go in [Discussions]( #community), not issues.\n\nEvery issue references the relevant whitepaper section via `ARCHITECTURE.md`, so you can see\nwhere a task sits in the whole before picking it up.\n\n## Community\n\n- **Discussions:** \u003chttps://github.com/name970/Protocol/discussions\u003e\n- **Discord:** https://discord.gg/aVwCK8WEYd\n\n## Cryptography, honestly\n\nThis is research, and the soundness arguments are being worked through in the open rather than\nasserted. Public discussion so far:\n\n- Hash-based nullifier without an elliptic-curve key tree'\u003chttps://crypto.stackexchange.com/posts/119735/revisions\u003e'\n- Commit-reveal hash binding sufficient to prevent adaptive forgeries in a Stochastic Matrix Audit\u003chttps://crypto.stackexchange.com/posts/119728/revisions\u003e\n \n\nIf you can break something, open an issue — adversarial review is the most useful contribution.\n\n## A note on the token\n\nThe broader protocol design includes a token. **It is not the point of this repository.**\nThis repo is about the technology; issues and discussion here are technical.\n\n## License\nMIT LICENSE\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fname970%2FProtocol","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fname970%2FProtocol","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fname970%2FProtocol/lists"}