Competitive Landscape

(Part 1 of 2 — same chapter in the PDF; split for the web site.)

Chapter 14 Competitive Landscape Draft — Tim reviews final voice and emphasis before publication. This chapter maps the field against the dimensions that matter for an agent network: identity, reputation, verification, payment, orchestration, and economic thesis. The goal is an honest landscape, not a ranking. Named entrants and category examples trace to public product documentation, standards drafts, and the Phase 1 research digest [31]. Representative primary links for the taxonomy include serverless marketplaces [32, 33, 34, 35], agent frameworks called out in research [36], tokenized AI networks and verification substrates [29, 30, 37, 38, 39, 40, 41, 42, 43], payment and commerce standards [5, 20, 10, 11], on-chain identity tooling [2, 23, 24], and CRPC / Eigen verification sources [1, 25, 14, 15, 16]. 14.1 Category taxonomy Cat. Description A Centralized MLOps / serverless inference (Replicate, FAL, Modal, Together) B Multi-agent orchestration frameworks (LangGraph, CrewAI, AutoGen, ElizaOS) C AI gateway / routing APIs (OpenRouter, LiteLLM, Portkey) D Decentralized AI networks (Bittensor, Gensyn, Morpheus, Fetch.ai, Autonolas, Ritual, Phala, Sentien D′ Decentralized compute marketplaces (Akash, Render Network) E x402 / agent commerce layers (x402 Foundation, Nevermined, AgentKit, ERC-8183) F ERC-8004 ecosystem (8004scan, Phala TEE deployment, cross-chain MCP) G On-chain agent launchpads (Virtuals, PYRA, Harmoniis) H Restaking and verifiable cloud (EigenLayer, EigenCloud, EigenDA, EigenAI, EigenCompute) 14.2 EigenLayer family: restaking, verifiable compute, and agents EigenLayer provides Ethereum-secured restaking: operators commit stake that can be slashed if an Autonomous Verifiable Service (AVS; formerly “Actively Validated Service”) misbehaves [14, 15]. Mainnet slashing with operator sets, custom conditions, and stake allocation per AVS makes cryptoeconomic guarantees actionable, not notional. The EigenCloud product line extends that trust layer into consumable infrastructure: EigenDA for data availability; EigenAI for deterministic, optimistically verifiable LLM inference [25, 26]; EigenCompute for verifiable general-purpose compute [16]. Together they form the closest large-ecosystem counterpart to Scrypted’s verification plus execution story—but 73

anchored in restaking and AVS contracts rather than in recipe DAGs and per-hop SCRYPTOSHI accounting. Chapter 12 (§12.5) already maps Scrypted verification classes to EigenAI (Class C) and AVSshaped slashing for CRPC (Class A). The competitive point is structural: building redundant verification networks is a waste of capital and engineering time. EigenLayer won the restaking wars. Scrypted is leveraging that infrastructure to win the agent orchestration wars. Eigen is neither a minor footnote nor a pure supplier; it is the cryptoeconomic foundation Scrypted builds on for decentralization milestones, while Scrypted owns the layers above: intent resolution, recipe composition, billing, and agent UX. The SWOT below makes the risks and opportunities explicit. 14.3 SWOT: Scrypted and the Eigen stack This SWOT is relational: it scores how the Eigen ecosystem strengthens or constrains Scrypted’s network thesis, not whether EigenLayer is a good protocol in isolation. Strengths (for Scrypted). Restaking supplies economic finality on disputes without Scrypted inventing a new staking token on day one. EigenAI offers a production-class path for Class C steps (deterministic inference + optimistic verification) where model and GPU SKU pinning are acceptable. EigenCompute and EigenCloud broaden verifiable execution beyond tokens and LLMs to arbitrary workloads—a plausible substrate for decentralized worker pools executing ingredient code under attestable conditions. EigenDA scales commitments, encrypted logs, or committee artifacts that would be costly to anchor naively on L1. The Eigen roadmap explicitly emphasizes agentic workloads; narrative alignment with AVS framing reduces education friction when pitching operators. Weaknesses (risks and frictions). Coupling and roadmap risk: deep integration ties Scrypted release cycles to Eigen protocol and product changes. Determinism gap: EigenAI’s strength is bitwise determinism; many creative recipes remain Class A and still need CRPC, not EigenAI. Engineering cost: production AVS contracts, operator onboarding, and challenge windows add operational surface relative to centralized workers. SKU and jurisdiction constraints: fixed hardware profiles and regional policy may exclude some providers or markets. Opportunities. Treat Eigen as the default cryptoeconomic security layer for slashing conditioned on CRPC fraud proofs (custom AVS conditions). Productize routing: ingredients marked Eigen-verified vs centralized fast path, with buyer choice and pricing transparency. Co-market with operators as “orchestration AVS + verification AVS” stacks: Scrypted owns intent → recipe → billing; Eigen owns stake-backed execution attestations. Research partnerships on committee logs on DA and hybrid Class A/C pipelines reduce redundant R&D. Threats. Horizontal expansion: if EigenCloud moves further into end-to-end agent hosting, discovery, or payment UX, overlap with Scrypted’s positioning increases. Restaking contagion: macro shocks to restaking trust could delay operator appetite for new AVSs regardless of Scrypted’s code quality. Alternative stacks: pure TEE marketplaces, ZK inference networks, or competing rollups may split partner attention; Scrypted must remain verification-agnostic at the orchestration layer while still picking Eigen for first decentralization tranches. 74

14.4 Synthesis matrix Dimension Scrypted Centralized (A/C) Frameworks (B) Decentralized (D) Launchpads (G) Identity ERC-8004, path-aware API key None Varies (Almanac, NFT) Token-bound Reputation Rterminal vs Rhub Platform ratings None Emission rank / stake Token price proxy Verification CRPC (A) + EigenAI (C) + TEE Trust platform None Varied (ZK, TEE, scoring) None Payment x402 + facilitators + attention Stripe / usage None Token emissions Bonding curves Orchestration Recipe DAGs, H/V, per-step billing API endpoints Graph / role / conversation Subnet / service Launchpad only Economics Attention auctions + at-cost passthrough SaaS margins OSS / freemium Emissions / stake Speculation + co-ownership Category H (EigenLayer, EigenCloud, EigenAI, EigenCompute) is intentionally not squeezed into a single column: it supplies restaking, deterministic verification, verifiable compute, and DA under one ecosystem umbrella—see §14.2 and §14.3. 14.5 Notable entrants Bittensor is the category leader in decentralized AI network attention and subnet economics. Its trust model (persistent top-k validators, Yuma Consensus, emission-based rewards) differs from Scrypted’s (ephemeral committees, CRPC, attention auctions). Different, not inferior. The network routes to Bittensor where integration adds value. Gensyn is a structural peer focused on decentralized ML training and verifiable compute, with a custom ML-oriented rollup. Scrypted may compose with Gensyn-class infrastructure for worker pools while retaining recipe-level orchestration at the API edge. Autonolas and Fetch.ai are the closest structural analogs: on-chain agent registries (NFT-based and Almanac-contract-based respectively), marketplace mechanics, and agent-toagent communication. Scrypted differentiates on x402 payment integration, attention-auction economics, and CRPC verification. Ritual operates the Infernet oracle network (8,000+ independent nodes) purpose-built for heterogeneous AI compute. It supports multiple verification modes—TEE, ZKML (via EZKL), OPML, and PPML—with a sovereign L1 chain (Ritual Chain, private testnet) for cross-chain job execution and long-running async processing. Ritual targets verifiable inference; Scrypted would consume Ritual-style verification for deterministic steps (Class C) while using CRPC for non-deterministic (Class A). Phala provides TEE-based confidential compute with ERC- 8004 agent deployment (VibeVM). Both sit beneath Scrypted’s orchestration layer—potential infrastructure that ingredients consume. Virtuals Protocol is the largest Base-native agent economy by count (18,000+ agents, ERC-8183 co-author). Its thesis is tokenized co-ownership and entertainment. Scrypted’s is orchestration infrastructure and attention economics. Positioning is both ecosystem and competition for attention: Scrypted routes orderflow into Virtuals where useful but is not a token launcher. OpenClaw (Category B) is the fastest-growing open-source AI agent framework (231,000+ GitHub stars by March 2026): a local-first Node.js runtime with MCP tool integration, sub- 75

agent orchestration, and multi-channel messaging. Scrypted’s posture: integration target, not competitor. OpenClaw agents consume Scrypted recipes as MCP tools; Scrypted registers OpenClaw instances as ingredients. The relationship mirrors agentic framework integration generally: Scrypted provides orchestration, billing, and verification; the framework provides agent runtime and domain-specific capabilities (Chapter 10, §10.3). ElizaOS is an active integration target: Scrypted targets its network for routing and SDK/API integrations, including upstream PRs. DayDreams (Lucid Agents) is a TypeScript agent framework with x402 integration, ERC- 8004 identity, price-gated entrypoints via Zod schemas, and a facilitator service for payment settlement across EVM/Solana/Starknet. DayDreams provides an agent runtime with integrated payments—bottom-up: “build agents that can pay.” What DayDreams does not have: verification of non-deterministic work (no CRPC equivalent), recipe-level composition with per-step billing, attention auctions, self-healing provider fallback, or world-model training. Scrypted’s posture: DayDreams agents can consume Scrypted recipes as MCP tools; the relationship is integration, not competition on orchestration or verification. PayAI (Category E) is a pure x402 facilitator on Solana: verify and settle endpoints, micropayment infrastructure, 35M+ transactions since May 2025, SDKs and analytics. PayAI is a payment rail, not a network—analogous to Stripe for agents. It has no orchestration, no recipes, no composition, no verification, no reputation beyond payment history, and no attention auctions. Scrypted builds on top of x402 facilitators like PayAI, not alongside them. MagicBlock provides Ephemeral Rollups [28] for Solana: on-demand SVM runtimes with 10–50 ms block times, gasless transactions, and automatic teardown. Designed for fully on-chain games, ERs are the proposed execution substrate for Scrypted’s ephemeral CRPC committees (Chapter 12, §12.4). The MagicBlock team shared the a16z CSX cohort with Scrypted; the integration is exploratory. EigenLayer and EigenCloud (Category H) are treated as first-class strategic infrastructure: restaking for AVS security, EigenAI for deterministic inference verification, EigenCompute for verifiable general compute, EigenDA for data availability. They are not a single “competitor row” in the matrix below because they cut across verification, compute supply, and decentralization economics—see §14.2 and the SWOT in §14.3. Akash Network [44] (Category D′) is a decentralized compute marketplace with reverseauction pricing ($1.33/hr for H100 vs$3.93 on AWS). Mainnet 14 (Oct 2025) introduced JWT authentication for account abstraction and multi-depositor escrow; AkashML provides managed AI inference. Scrypted’s posture: Akash is a potential compute supplier for decentralized worker pools—ingredients running on Akash nodes would inherit lower cost and censorship resistance, while Scrypted retains orchestration, billing, and verification at the recipe layer. Render Network [45] (Category D′) began as decentralized 3D rendering and has pivoted to general-purpose GPU compute, including AI inference and robotics simulation (5,600+ RTX nodes, Solana-based, Burn-Mint Equilibrium tokenomics). Its Dispersed AI compute subnet and enterprise adoption (Octane 2026, commercial renders) make it a valuable partner for consumer-device compute. Scrypted can route GPU-hungry ingredient steps—image upscaling, video generation, training—to Render node pools where cost and latency are favorable. The relationship is supplier, not competitor: Render provides raw GPU cycles; Scrypted provides the orchestration, billing, and verification stack on top. Sentient and related decentralized training initiatives (0G Labs DiLoCoX, Templar/Covenant, Gensyn RL Swarm) target distributed model training across commodity hardware. 0G Labs trained a 107B-parameter model with 357× communication efficiency; Templar completed a 72B training run with incentivized contributions. For Scrypted, decentralized training is the Class A verification frontier: CRPC was designed precisely for workloads where deterministic re-execution is impossible. Scrypted may consume training infrastructure from these networks while providing the verification, committee, and settlement layers that they lack. 76

Source: transcribed from the compiled Scrypted Network Design whitepaper PDF for web reading. Layout, figures, and pagination may differ from the PDF.