<div id="readability-page-1"><div><div><p>Author: Kevin, Movemaker Researcher; Source: X, @MovemakerCN</p><h2>Introduction: A Structural Leap from Generative AI to “Agent Behavior”</h2><p>In 2026, the field of artificial intelligence will experience a structural leap from “generative capabilities” to “agent action capabilities.” If 2023-2024 was about the amazing language generation capabilities of large language models, then 2026 will mark the formal establishment of the “AI Agent Economy.”</p><p>Based on the predictions and analysis of the a16z Crypto research team, further research reveals that 2026 will be a year of deep integration between AI, a productivity tool, and Crypto, a value distribution layer.</p><p> AI is no longer merely a passive tool responding to human commands, but an active participant with reasoning, planning, trading, and autonomous discovery capabilities. Based on a16z Crypto's outlook report, the three core trends reshaping the AI+Crypto landscape in 2026 are: A new research paradigm: from single-agent to &#34;Agent-Wrapping-Agent&#34;. A revolution in financial infrastructure: from KYC to KYA (Know Your Agent). Economic model reconstruction: addressing the &#34;hidden tax&#34; crisis facing open networks through payment and programmable IP. These three trends are not isolated: the shift in research paradigms relies on advanced collaboration between agents; advanced collaboration requires agents to possess verifiable identities (KYA); and agents with identities must adhere to new value exchange protocols when acquiring data. 1. The New Era of Polymaths: The &#34;Agent-Wrapping-Agent&#34; Architecture in Advanced Research Starting this year, the definition of &#34;AI-assisted research&#34; will undergo a qualitative leap. We will no longer talk about simple literature retrieval or text summarization, but witness AI systems capable of substantive reasoning, hypothesis generation, and even autonomously solving doctoral-level problems. The core driving force behind this transformation lies in the shift from linear suggestion engineering of a single model to complex, recursive AWA workflows. 1.1 Breakthrough in Reasoning Ability: Crossing the Boundaries of Pattern Matching Scott Kominers of a16z points out that AI models are evolving from simply understanding instructions to being able to receive abstract instructions (like guiding a doctoral student) and return novel and correctly executed answers. Recent technological advancements indicate that AI models are breaking through the &#34;random parrot&#34; ceiling, exhibiting slow, deliberate reasoning abilities similar to human &#34;systematic&#34; thinking. 1.1.1 The &#34;Illusion of Usefulness&#34; With the enhancement of reasoning abilities, a new style of &#34;polymath&#34; research is emerging. Scott describes this style as: &#34;Using AI to cross disciplinary boundaries, speculating on potential deep connections between topology and economics, biology and materials science.&#34; The &#34;illusionary&#34; nature of large models, often criticized, is being reconstructed as a &#34;generative exploration&#34; mechanism within the context of scientific discovery: Protein Design Case Study: Researchers at the University of Washington used the &#34;whole-family illusion&#34; concept to generate over 1 million unique protein structures not found in nature. Among them, a novel luciferase was screened that exhibited catalytic activity comparable to natural enzymes but with higher substrate specificity. Fluid Dynamics Discovery: Through Physical Information Neural Networks (PINNs), researchers discovered new unstable singularities in Navier-Stokes equations, revealing previously unknown patterns in fluid motion. The core of this research style lies in allowing the model to &#34;think wildly&#34; in the abstract space to generate high-entropy conjectures, which are then filtered using a rigorous logical verifier. 1.2 Detailed Explanation of AWA Architecture To harness this powerful reasoning and generative capability, research workflows are shifting from flattened to hierarchical structures. AWA refers not only to dialogue between multiple agents but also to a recursive, hierarchical control structure. 1.2.1 Orchestrator-Executor Pattern This is currently the most mainstream AWA implementation pattern. A &#34;principal researcher&#34; agent is responsible for maintaining the global context and research objectives, decomposing tasks and distributing them to a dedicated group of &#34;executor&#34; agents. <strong>Architectural Advantages</strong>: Data from Anthropic shows that a multi-agent system consisting of Claude Opus as the master agent and Claude Sonnet as a sub-agent outperforms a single Claude Opus agent in complex research tasks by 90.2%. This performance improvement is mainly attributed to context isolation—the master agent does not need to process redundant information for each sub-task, thus maintaining the clarity of reasoning.</p><p><strong>1.2.2 Recursive Self-Improvement and the MOSAIC Framework</strong></p><p>Another key feature of the AWA architecture is the introduction of a Reflexion loop. When the underlying agent fails to execute a task, the error information is fed back to a &#34;critic&#34; agent for analysis and correction.</p><p>The MOSAIC framework (Multi-Agent System for AI-driven Code)</p><p> By introducing specialized &#34;self-reflection agents&#34; and &#34;principle generation agents,&#34; the accuracy of scientific code generation has been significantly improved without relying on verification test cases. This closed loop of &#34;trial and error-reflection-retry&#34; simulates the thought process of human scientists when facing experimental failures. 1.3 Case Study: Sakana AI's &#34;AI Scientist&#34; The most noteworthy AWA application case in 2025 is Sakana AI's &#34;The AI ​​Scientist&#34; system. This is a system designed to fully automate the entire lifecycle of scientific discovery. 1.3.1 Fully Automated Scientific Research Closed-Loop Process The system, based on a starting code template (such as NanoGPT), uses LLM as a &#34;mutation operator&#34; to brainstorm diverse research directions and calls the Semantic Scholar API. Literature review ensures novelty. Experimental Iteration: The &#34;Experimenter&#34; Agent writes and executes code. If the experiment fails, the system captures error logs using the Aider tool and autonomously repairs the code until a visualization is obtained. Paper Writing: The &#34;Writer&#34; Agent uses LaTeX to write a complete scientific paper, covering abstract, methods, experimental results, and autonomously finds cited literature to generate BibTeX. Automated Peer Review: The generated paper is submitted to a simulated &#34;Reviewer&#34; Agent, which scores it according to the standards of top conferences (such as NeurIPS). The system can even make multiple revisions based on reviewer comments. 1.3.2 Economic Benefits and Quality The economic efficiency of the &#34;scientist&#34; system is astonishing: the computational cost of generating a complete research paper is only about $15. The paper generated by this system, &#34;Compositional Regularization,&#34; even successfully passed peer review at the ICLR workshop. Although limitations such as citation illusions and logical flaws still exist, this case demonstrates that AI has the ability not only to assist research but also to execute the entire research process. 2. The Command of Identity: From KYC to KYA As agents are given the power to perform tasks and transactions, the digital economy faces an unprecedented identity crisis. Sean Neville (CEO of Catena Labs) warns that the number of &#34;non-human identities&#34; in the financial services sector has reached 96 times the number of human employees, and even as high as 100:1 in some statistics. These agents—without bank accounts, without real-name authentication, yet operating at machine speed—are a huge compliance black hole. The industry is urgently shifting from traditional KYC to... KYA (Know Your Agent). 2.1 The Emergence and Risks of Non-Human Identities (NHI) 2.1.1 “Shadow AI” and the 96:1 Imbalance 45% of financial service institutions acknowledge the existence of unapproved “shadow AI agents” within their organizations. These agents create “identity silos” outside the formal governance framework. Risk Scenarios: A test agent used for cloud resource optimization might autonomously purchase expensive reserved instances without human intervention; or a trading bot might trigger erroneous sell orders during market volatility. Attribution Dilemma: When an agent violates regulations, who is responsible? The engineer who developed it? The manager who deployed it? Or the vendor who provided the underlying model? Without KYA, these responsibilities cannot be defined. 2.2 KYA Framework: The Trust Foundation of the Machine Economy KYA is not just about issuing ID cards, but about establishing a complete digital identity system encompassing subjects, credentials, permissions, and reputation. 2.2.1 The Three Pillars of KYA alt=&#34;jR2oyBeNftkjgbuN2nXhBfJyvRnuLBJu5BFcacmH.jpeg&#34;&gt;</p><ol><li>Subject: The entity legally responsible for the Agent. The agent must be cryptographically linked to a KYC/KYB verified human or business account.</li><li>Agent Identity: A unique digital identity based on a decentralized identifier. The DID is cryptographically generated, immutable, and cross-platform portable.</li><li>Mandate/Delegation: A statement of authority issued through Verifiable Credentials (VCs). For example, a VC could state: &#34;This Agent is authorized to make purchases on behalf of Alice on Amazon, with a maximum limit of...&#34; &#34;$500&#34;.</li></ol><p><strong>2.2.2 Cryptographic Binding and Trust Chain</strong></p><p>When an Agent initiates a transaction, it presents a VC. The verifier does not need to trust the Agent itself, but only needs to verify whether the digital signature on the VC comes from a trusted issuer. This mechanism creates a &#34;trust chain&#34;: Bank trusts enterprise -&gt; Enterprise issues VC to Agent -&gt; Merchant verifies VC -&gt; Transaction passes.</p><h3>2.3 Protocol Stack Debate: Standardization of Agent Identity</h3><p><strong>2.3.1 Skyfire and KYAPay Protocol</strong></p><p>Skyfire launched the KYAPay open standard, whose core innovation lies in composite tokens:</p><ul><li>kya token: Contains identity information (such as &#34;verified enterprise agent&#34;).</li><li>pay token: Contains payment capability (such as &#34;pre-authorization of 10 USDC&#34;).</li><li>kya+pay token: Packages identity and payment, allowing the agent to complete &#34;visitor checkout&#34; without manual form filling.</li></ul><p><strong>2.3.2 Catena Labs and ACK (Agent Commerce Kit)</strong></p><p>Catena Labs, founded by USDC architect Sean, launched ACK, aiming to create &#34;HTTP for intelligent agent commerce&#34;. ACK emphasizes the use of W3C DID.</p><p> Standards and account abstractions allow agents to directly control on-chain smart contract wallets, achieving stronger security than API keys. 2.3.3 Google AP2 and x402 Extensions Google's Agent Payments Protocol (AP2) uses &#34;authorization letters&#34; to manage permissions and has partnered with Coinbase to develop the AP2 x402 extension, directly integrating crypto payment standards into the protocol. 2.4 Smart Agent Credit Scoring and Risk Control KYA is still the beginning of a reputation system. On-chain Reputation (ERC-7007): Through ERC-7007 (verifiable AI) (Generate content token standards), every successful interaction of the Agent (such as timely payment, generating high-quality code) can be recorded on the chain, forming a verifiable history. Real-time circuit breakers: Financial institutions are deploying AI gateways. If the behavior of a transaction Agent deviates from the benchmark (such as high-frequency abnormal trading), the system can immediately revoke its VC, triggering &#34;digital suppression.&#34; 3. Economic Restructuring: Solving the &#34;Hidden Tax&#34; of Open Networks Liz of a16z points out that AI Agents are levying a &#34;hidden tax&#34; on open networks: In order to serve users, Agents massively extract data (context layer) from content websites, but systematically bypass the advertising and subscription models that support the production of this content. If this parasitic relationship is not resolved, it will lead to the depletion of the content ecosystem. 3.1 &#34;The Great Decoupling&#34;: The Full Arrival of the Zero-Click Economy In 2025, the digital publishing industry witnessed the &#34;Great Decoupling&#34;: search volume rose, but the number of clicks flowing to websites plummeted. 3.1.1 The Brutal Data of Traffic Erosion [Image of a file: https://img.jinse.com.cn/7427173_watermarknone.png] Zero-Click Rate Soars: a16z predicts that traditional search engine traffic will decline by 2026. 25%. Similarweb data shows that the zero-click search rate has risen to 65% in 2025. </p><li>Click-through rate (CTR) collapse: DMG Media reports that when AI Overview appears above search results, the click-through rate of its content plummets by 89%. Even the top-ranked search result loses 34.5% of clicks in front of AI Summary. </li><h3>3.2 Moving away from static licensing: A new pay-as-you-go model</h3><p>To cope with this crisis, the industry is shifting from static annual data licenses (such as the deal between Reddit and OpenAI) to pay-as-you-go compensation. </p><p><strong>3.2.1 Perplexity's Comet Plus model</strong></p><p>Perplexity AI launched...</p><p> Server returns 402 Payment Required with a price (e.g., 0.001 USDC) -&gt; Agent automatically signs the payment via L2 blockchain (e.g., Base, Solana) or Lightning Network -&gt; Server verifies and releases the data. Economic Efficiency: Traditional payment gateways cannot handle transactions worth a few cents, while x402 combined with a low-fee blockchain reduces costs to negligible levels, making payment possible. 3.3.2 Machine-Readable Rights: TDMRep and C2PA left;&#34;&gt;<strong>TDMRep (Text Data Mining Reserved Protocol): A W3C community standard that allows websites to declare in robots.txt or HTTP headers: &#34;Reserved TDM rights, payable/licensed.&#34; This provides a clear binary signal for the Agent.</strong></p><li>C2PA (Content Origin and Authenticity Alliance): Proves the original source of content by embedding tamper-proof &#34;content credentials.&#34; Even if content is ingested by AI, the cryptographic signature provided by C2PA ensures that the attribution chain remains unbroken, providing a basis for royalty distribution.</li><h3><strong>3.4 On-Chain IP Attribution: Story Protocol</strong></h3><p><strong>A more radical change is the tokenization of intellectual property itself. Story Protocol is committed to building a &#34;programmable IP&#34; layer.</strong></p><ul><li></li></ul></div></div></div>

<p>The 2026 AI Agent Economic Outlook predicts a significant shift from generative AI to agent action capabilities, marking the establishment of the "AI Agent Economy." Key trends include a new research paradigm of "Agent-Wrapping-Agent," a revolution in financial infrastructure from KYC to KYA, and economic model reconstruction to address hidden tax crises in open networks. The report highlights advancements in AI reasoning abilities, enabling complex research tasks and interdisciplinary exploration, exemplified by breakthroughs in protein design and fluid dynamics. The AWA architecture enhances research workflows through multi-agent collaboration and self-improvement mechanisms.</p>

2026 AI Agent Economic Outlook: Reshaping AI Identity and the Flow of Network Value