Crynux (crynux.io) positions itself as a decentralized AI orchestration layer running on edge devices, with its native token CNX deployed on Base (EIP-155:8453). Its official website describes a multi-stage roadmap—Hydrogen (consensus), Helium (inference), Lithium (fine-tuning/training), and upcoming Beryllium (workflow and tokenization)—and reports live network metrics including 2,170 TFLOPS of computing power and over 235,000 tasks executed. The CNX token contract is verified on BaseScan at 0x9557…9396, and technical documentation is published via ResearchGate. According to API-sourced claims, the total supply is 8.62 billion CNX, with a self-reported circulating supply of ~592 million; however, the same source states ‘0 in circulation’—a material discrepancy requiring reconciliation. No treasury balance, reserve composition, or funding runway is disclosed on the website or in the supplied evidence. Critical evaluation questions remain: What assets back the token? Where are treasury funds held, and under what custodial control? Is there verifiable evidence of revenue generation, staking yield sustainability, or off-chain commitments supporting the stated tokenomics?
Crynux presents a detailed technical vision for decentralized AI inference and training on edge devices, anchored in its Hydrogen consensus protocol and modular roadmap (Hydrogen → Helium → Lithium → Beryllium). The website crynux.io describes live metrics—2,170 TFLOPS, 217 nodes online, 235,499 tasks executed—and promotes production-ready API access, node operation, and staking. However, no evidence confirms these metrics are live or verifiable: the page is static, with no interactive demo, API playground, or observable user journey (e.g., no sign-up flow, no working endpoint test, no authenticated dashboard). The technical documentation points to a ResearchGate paper, but that publication does not demonstrate functional software or runtime validation. Token data from API sources reports zero circulating supply and stale last-updated timestamps (July 2026), raising questions about economic activation. All claims—including consensus overhead (~10%), Sybil resistance, and multi-modality support—are asserted without third-party verification, audit references, or open-source repository links (the provided BaseScan URLs point to token pages, not code). No evidence shows repeat usage, retention cohorts, or failure handling. Because the primary thesis—that Crynux delivers a working, trustless, decentralized AI compute layer—is unverified in the supplied snapshot, the assessment is capped at 5 per rubric hard cap.
Overall score: 5/10 Confidence: Low