{"id":"d219500f-7570-4da6-8c69-305353ebcedb","arxiv_id":"2606.24942","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":3.0,"correctness_risk":"high","formal_verification":"none","parameter_count":1,"one_line_summary":"Proposes Proof-of-Useful-Work for AI tasks in a token economy with post-quantum security, formalized via a closed-loop model and sufficient-stake condition.","lead":"The paper proposes a three-layer architecture for a decentralized AI economy that rewards nodes for performing machine learning inference and training instead of hash-based Proof-of-Work. It claims this approach provides both economic value and better resistance to quantum attacks on signatures via post-quantum cryptography.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.3","headline":"Decentralized verification of ML linear algebra lacks any specified protocol or attack-surface analysis","rationale":"The reader's weakest_assumption already isolates exactly the verification and stake-condition gaps that the abstract's architecture description leaves unaddressed; the quantum-resistance half of the claim is comparatively well-supported by the Grover/Shor distinction and post-quantum migration, so the load-bearing uncertainty remains on the ML-verification side.","tokens_in":1643,"tokens_out":316,"duration_ms":13024,"concrete_test":"Provide an explicit verification protocol for a single linear-algebra primitive (e.g., matrix-vector multiply) together with its communication and computation cost relative to SHA-256; if the protocol either exceeds 5\times the cost of a hash or admits a forgery attack with probability >2^{-40} under the stated threat model, the advantage claim does not hold.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that ML-native operations (training/inference linear algebra) can be verified efficiently by the validation layer without new attack surfaces or prohibitive overhead, and that the derived sufficient-stake condition in the (\theta_c, \theta_w, W)-closed-loop economy actually deters dishonesty. The three-layer architecture is named but no concrete verification primitive (e.g., interactive proof, zk-SNARK for matrix multiplication, or redundancy scheme), no overhead bound, and no game-theoretic analysis of the stake condition are supplied; without these the economic-advantage half of the claim has no demonstrated foundation.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes a decentralized AI economy based on Proof-of-Useful-Work (PoUW) where nodes perform machine learning inference and training tasks instead of hash-based puzzles. It introduces a three-layer architecture (compute, validation, economic coordination), formalizes the system as a (θ_c, θ_w, W)-closed-loop token economy from which a sufficient-stake condition for honest participation is derived, and argues that this approach provides both economic value and quantum resilience compared to classical Proof-of-Work, since Grover's algorithm yields only quadratic speedup on hashes but not on ML linear algebra, while post-quantum cryptography can mitigate Shor's algorithm threats to signatures.","tokens_in":1785,"tokens_out":532,"duration_ms":19026,"significance":"If the missing technical details were supplied and the claims substantiated, the work would be significant for advancing sustainable blockchain consensus mechanisms that produce external value through useful computation while addressing emerging quantum threats. The integration of ML workloads with post-quantum security primitives represents a novel direction in the field.","major_comments":[{"comment":"Abstract: The sufficient-stake condition is stated to be derived from the (θ_c, θ_w, W)-closed-loop token economy, but no equations, proofs, or validation data are provided to support this derivation or to demonstrate that it enforces honest behavior externally rather than by construction.","section":"Abstract"},{"comment":"Validation layer description: No specific protocol, interactive proof, zk-SNARK construction, or redundancy scheme is given for verifying ML linear algebra operations in a decentralized setting, nor is any overhead analysis or attack-surface evaluation supplied; this omission is load-bearing for the claimed economic advantages.","section":"Validation layer description"},{"comment":"Quantum advantage paragraph: The claim that Grover's algorithm does not accelerate ML-native linear algebra is asserted in a single sentence without supporting analysis, complexity bounds, or references, weakening the quantum-security advantage argument.","section":"Quantum advantage paragraph"}],"minor_comments":[{"comment":"Abstract: Minor grammatical issues, e.g., 'instead of ineffective hashing method' should be 'instead of the ineffective hashing method'.","section":"Abstract"},{"comment":"Notation: The parameters θ_c, θ_w, W are introduced without explicit definitions or ranges in the abstract, which may confuse readers.","section":"Notation"}],"recommendation":"major_revision","confidential_remarks":"The manuscript appears to be a high-level proposal without the detailed technical development expected in cs.CR; consider whether it fits the journal's standards for rigor in security and cryptography submissions."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback. We address each major comment below and indicate planned revisions to provide the requested technical details and strengthen the arguments.","responses":[{"response":"The manuscript presents the closed-loop model conceptually. We will revise to include the explicit equations for the (θ_c, θ_w, W) parameters, the full derivation of the sufficient-stake condition, and a proof sketch showing enforcement of honest behavior through economic incentives external to the model assumptions.","revision_made":"yes","referee_comment":"[Abstract] Abstract: The sufficient-stake condition is stated to be derived from the (θ_c, θ_w, W)-closed-loop token economy, but no equations, proofs, or validation data are provided to support this derivation or to demonstrate that it enforces honest behavior externally rather than by construction."},{"response":"We agree this detail is essential. The revised manuscript will specify a redundancy scheme with zk-SNARKs for verifying the linear algebra operations, include overhead analysis, and evaluate the attack surface to substantiate the economic advantages of the validation layer.","revision_made":"yes","referee_comment":"[Validation layer description] Validation layer description: No specific protocol, interactive proof, zk-SNARK construction, or redundancy scheme is given for verifying ML linear algebra operations in a decentralized setting, nor is any overhead analysis or attack-surface evaluation supplied; this omission is load-bearing for the claimed economic advantages."},{"response":"The claim follows from the quadratic speedup of Grover's algorithm applying only to unstructured search, not to the structured linear algebra central to ML workloads. We will expand the paragraph with supporting analysis, complexity bounds, and references to quantum algorithms for linear systems and ML to reinforce the quantum-resilience argument.","revision_made":"yes","referee_comment":"[Quantum advantage paragraph] Quantum advantage paragraph: The claim that Grover's algorithm does not accelerate ML-native linear algebra is asserted in a single sentence without supporting analysis, complexity bounds, or references, weakening the quantum-security advantage argument."}],"tokens_in":1341,"tokens_out":451,"duration_ms":24637,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper's main move is to replace hash-based PoW with ML training and inference as the work that earns tokens, while adding post-quantum signatures. It correctly observes that Grover's algorithm gives only quadratic speedup on hashes and that Shor's algorithm hits classical signatures, so lattice or hash-based signatures are the right fix on that side. The three-layer split (compute, validation, economic coordination) and the (\theta_c, \theta_w, W) closed-loop parameters are the only named formal elements.\n\nNothing else is delivered. The abstract claims a sufficient-stake condition is derived, yet no equations, game, or bounds appear. The hard part—how the validation layer checks ML linear algebra without new attack surfaces or prohibitive cost—is not addressed at all; no interactive proof, zk primitive, redundancy scheme, or overhead estimate is given. Because the token economy is defined as closed-loop, any honesty property derived from it stays internal and does not rest on external benchmarks.\n\nThe quantum-security half is standard and already known; the economic-advantage half has no demonstrated foundation once verification is required. The work is therefore an architectural outline rather than a technical result.\n\nIt is aimed at readers who follow blockchain-AI intersections at the level of proposals and roadmaps. Anyone looking for verifiable protocols, reproducible measurements, or formal arguments will find the paper empty on those points. It does not reach the threshold for serious refereeing.","headline":"This is a high-level sketch that names a three-layer architecture and a token-economy parameterization but supplies none of the verification protocols, proofs, or overhead analysis needed to make the claims hold.","tokens_in":2286,"tokens_out":373,"would_cite":false,"duration_ms":12117,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Replacing hash puzzles with verified machine learning work yields both economic productivity and quantum security in decentralized networks.","keywords":["proof-of-useful-work","decentralized AI economy","post-quantum security","blockchain consensus","machine learning verification","token economy model"],"falsifier":"A demonstration that decentralized verification of ML linear algebra either incurs prohibitive computational overhead or creates exploitable vulnerabilities, or that no choice of token economy parameters satisfies the sufficient-stake condition against rational dishonest actors.","tokens_in":2535,"feed_emoji":"🔐","tokens_out":679,"duration_ms":18698,"temperature":0.7,"pith_summary":"The paper proposes a decentralized AI economy that rewards nodes for performing machine learning inference and training rather than solving hash puzzles. It introduces a three-layer architecture for compute, validation, and economic coordination, formalized through a closed-loop token economy with parameters (θ_c, θ_w, W) that derives a sufficient-stake condition for honest behavior. Because Grover's algorithm provides only quadratic speedup against hashes but not against ML linear algebra, and post-quantum cryptography can secure signatures against Shor's algorithm, this useful-work approach offers advantages in both economics and quantum resilience over traditional proof-of-work.","feed_headline":"ML computations replace hash puzzles in quantum-safe consensus","feed_subtitle":"A three-layer architecture rewards useful AI work while deriving stake conditions that resist Grover and Shor attacks through post-quantum s","key_machinery":"The three-layer architecture separating compute, validation, and economic coordination, combined with the (θ_c, θ_w, W)-closed-loop token economy that enforces a sufficient-stake condition.","core_discovery":"Useful-work consensus, where nodes are rewarded for machine-learning inference and training, is implemented via a three-layer architecture that separates compute, validation, and economic coordination; this is formalized by a (θ_c, θ_w, W)-closed-loop token economy that derives a sufficient-stake condition for honest participation, providing both economic value and resistance to quantum attacks since Grover's algorithm does not accelerate ML-native linear algebra while post-quantum standards mitigate Shor's threats to signatures.","pith_inferences":["Verification of ML operations could extend to other verifiable computations like scientific simulations.","Parameter tuning in the closed-loop economy may require real-world testing to balance incentives.","Adoption could lead to decentralized marketplaces for AI model training and inference.","This approach might reduce the environmental impact of blockchain consensus by tying rewards to useful outputs."],"forward_implications":["Machine learning computations produce external economic value instead of wasted energy on hashing.","The system resists quadratic speedup from Grover's algorithm on hash puzzles.","Post-quantum migration to lattice-based and hash-based signatures protects against Shor's algorithm.","The sufficient-stake condition in the token economy encourages honest participation.","Useful-work consensus provides both economic and quantum-security advantages over classical proof-of-work."],"fun_headline_variants":["Useful ML work replaces hash puzzles in blockchain consensus","Quantum-resilient economies reward AI training and inference","Post-quantum security via useful work in decentralized AI","Three-layer architecture enables quantum-safe useful-work consensus"],"cache_read_input_tokens":64,"weakest_assumption_plain":"Machine-learning linear algebra operations can be efficiently verified in a decentralized setting without introducing new attack surfaces or excessive overhead, and the closed-loop token economy parameters can be set such that the sufficient-stake condition enforces honest behavior.","fun_headline_variants_meta":{"raw":{"variants":["Useful ML work replaces hash puzzles in blockchain consensus","Quantum-resilient economies reward AI training and inference","Post-quantum security via useful work in decentralized AI","Three-layer architecture enables quantum-safe useful-work consensus"]},"model":"grok-4.3","cost_usd":0.008177,"raw_usage":{"total_tokens":3686,"prompt_tokens":615,"num_sources_used":0,"completion_tokens":58,"cost_in_usd_ticks":81774500,"prompt_tokens_details":{"text_tokens":615,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3013,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":615,"tokens_out":58,"duration_ms":20711,"temperature":1.0,"reasoning_tokens":3013,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-26T07:32:45.293328+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A demonstration that decentralized verification of ML linear algebra either incurs prohibitive computational overhead or creates exploitable vulnerabilities, or that no choice of token economy parameters satisfies the sufficient-stake condition against rational dishonest actors.","supporting_citations":[],"review_version":1}