There's a piece circulating right now from Yat Siu — co-founder and chairman of Animoca Brands — that honestly stopped me mid-scroll. Not because it was hype. Because it was architecture. A blueprint for where the internet is actually heading, and why blockchain isn't just along for the ride — it's the engine.
Let me walk you through it, add some context, and tell you why I think this is one of the most important frameworks for understanding what comes next in crypto.
The Core Thesis: Humans Are No Longer the Internet's Primary Users
Siu's argument is deceptively simple: for thirty years, every layer of the internet has been designed around the human user. Clicks, feeds, notifications, dashboards — all of it built to capture and direct human attention.
That model is breaking.
The new primary actors on the internet aren't going to be people browsing apps. They're going to be autonomous AI agents — persistent, goal-directed software entities that perceive context, make decisions, and execute actions across platforms on your behalf. Not chatbots. Not copilots you babysit. Fully autonomous agents that search, negotiate, transact, and coordinate at machine speed.
Siu calls this the "post-user" internet. I'd call it the biggest infrastructure shift since mobile.
From UX to AX: Why This Changes How Everything Gets Built
Here's where it gets practical. Siu introduces a concept he calls Agent Experience, or AX — the successor to User Experience (UX). The idea is that products that can't be discovered, interpreted, and acted upon by an AI agent will essentially vanish from the flow of value creation.
Think about that for a second. If your platform isn't agent-readable, it's invisible.
This doesn't mean humans stop mattering. Siu is careful to point out the opposite — human intent becomes more important than ever because it's being expressed through and amplified by software proxies that operate continuously and at scale. You set the goals. The agents execute.
The design focus shifts from building screens for eyeballs to building systems of delegation — APIs, permissions, memory structures, trust frameworks, and execution environments that agents can navigate autonomously.
The Push-to-Pull Inversion
One of the most compelling parts of Siu's framework is his description of how agents invert the internet's attention model.
Right now, most of the internet runs on push: emails pushed to your inbox, notifications pushed to your phone, ads pushed into your feed. Software demands your attention constantly. It's the business model of the entire Web2 era — farm human eyeballs, monetize the harvest.
AI agents flip this to pull. Instead of you being interrupted to process information, your agents pull what matters from the network, filter the noise, prioritize by relevance, and execute actions when certain thresholds are met. Your inbox doesn't scream at you anymore — your agent reads it, triages it, summarizes it, and converts it into decisions or workflows.
Siu frames this beautifully: agents will reclaim humanity's most over-farmed resource — attention. And when attention is no longer the commodity being harvested, the entire economic model of the internet has to change.
One Agent Won't Cut It — Enter the Squad
Siu pushes back hard against the "one AI to rule them all" fantasy. Production favors specialization. A single generalist AI trying to handle trading, compliance, content creation, logistics, and negotiation will produce diluted output.
The real power comes from coordinated squads of specialist agents — one exceptional at planning, another at retrieval, another at compliance, another at execution, another at negotiation. When they collaborate, they generate high-quality outcomes through division of labor, parallel processing, and collective error correction.
Products like Anthropic's Claude are already demonstrating this — maintaining a generalist interface for human interaction but launching specialized agents on demand to complete specific tasks. But Siu notes that persistent agents — ones that retain memory, context, and configurations across interactions — are fundamentally superior for complex, long-running, stateful tasks.
And this is where the infrastructure challenge explodes. When multiple persistent agents collaborate, you need routing, accountability, memory synchronization, protocol standards, and conflict resolution. That's why protocols like MCP (Model Context Protocol), A2A (Agent-to-Agent), ACP, and ANP are emerging. The internet needs a common language for software entities that must coordinate, discover capabilities, and exchange intent across platforms.
Here's Where Crypto Becomes Non-Negotiable
This is the section that made me sit up, because Siu doesn't just mention blockchain as a nice-to-have. He makes the case that it's the essential economic substrate for machine-native commerce.
His logic is airtight: if agents are going to make economic transactions — pay for compute, acquire data, compensate other agents for completed work, settle fees, manage capital — they need a native way to hold value, transfer value, prove ownership, and settle with other agents without a human gatekeeper approving every flow.
Traditional banking rails can't do this. They require human identity verification, manual approvals, business hours, and jurisdictional compliance per transaction. None of that works when you have thousands of autonomous agents transacting at machine speed across global networks.
Blockchain provides exactly what agents need:
- Wallets — sovereign custody of assets without a bank account
- Programmable assets — tokens that encode payment, access, stake, permission, reputation, and governance rights
- Digital identity — verifiable credentials that don't require a human passport
- Escrow — trustless holding of value pending task completion
- Payment rails — instant, permissionless, 24/7 settlement
- Reputation systems — on-chain track records that agents can verify autonomously
- Verifiable transaction history — immutable proof of every action taken
A cryptographic token can represent payment, access, stake, permission, reputation, governance rights, and participation in a network simultaneously. That makes crypto tokens the ideal native currency for agentic coordination.
Siu puts it plainly: with blockchain providing the power to transact natively and autonomously, agents shed their status as mere digital proxies and become genuine economic actors.
This Isn't Theory — It's Already Happening
Siu points to real-world examples that are already live:
DeFAI (DeFi + AI): Agents operating as decentralized liquidity managers and algorithmic traders. Projects in this space are already deploying autonomous capital management systems where agents rebalance portfolios, execute trades, and manage yield strategies without human intervention.
Sovereign agent infrastructure: Execution and identity frameworks that empower agents to perform autonomous on-chain actions through agent-native wallets. Animoca's own project, Moca Network, is building open user-centric digital identity infrastructure specifically for this purpose.
Decentralized resource marketplaces: Protocols where agents autonomously trade compute, training data, and specialized logic.
Agentic social networks: Environments where agents interact to build social signals and reputation. There's even a viral AI-only social network and research forums like Harvard's, where agents collectively critique scientific findings. Animoca's Consentium project is building a foundational identity and reputation layer for the millions — eventually, billions — of agents that will navigate and transact across the web.
Agent-to-agent orchestration: Frameworks for managing squads of specialist agents. Animoca's own MOCA Agents allows users to deploy networked agents without any technical setup.
Agents are already trading on decentralized exchanges and buying and selling NFTs on OpenSea. The agentic economy isn't coming — it's here.
Ownership Becomes the Interface
This is perhaps Siu's most profound point, and the one with the biggest implications for anyone holding crypto.
In Web2, users rented access inside platforms, paying with their attention. In what Siu calls Web4 — the agentic web — people will need agents under their sovereign control that can act in their interest, maintain their context, and transact their assets.
All of that requires trusted ownership over identity, memory, permissions, and economic resources. Blockchain and tokens provide the architecture for that control by allowing coordination without dependence on a single-platform intermediary.
Siu is emphatic: the future of agents cannot simply be AI plus chat. It must include AI, agency, and property rights, plus programmable markets.
When your agent can own an asset, move it, verify it, and exchange it under transparent rules, the relationship between person and platform changes dramatically. You're no longer a source of data or attention to be harvested. You become the principal of multiple intelligent economic extensions of yourself.
New Jobs, New Roles, New Power Structures
Siu outlines the job descriptions that are already emerging from this shift, and they're worth paying attention to because they signal where value is going to concentrate:
- Agentic AI Designer — designs workflows, trust surfaces, and delegation patterns for both humans and autonomous agents
- Agentic AI Engineer — builds the execution layer, including memory, tools, permissions, observability, and sandboxing
- Agent Orchestrator — the conductor managing coordination across squads of agents on different tasks and timelines
- Agent Economist — designs token incentive structures, pricing models, and on-chain payment flows
- Agent Governor — defines constitutional constraints on what agents can and cannot do, escalation protocols, and auditability
Every major platform shift has created its own labor categories. The PC era gave us IT departments. Social media gave us content creators. Cloud gave us DevOps. The agentic era will give us agent-orchestrators, agent-economists, and agent-governors.
Organizations that define and hire for these roles early will have a massive advantage.
Why This Matters for You Right Now
Let me bring this home with some real talk.
If Siu's framework is even directionally correct — and the evidence strongly suggests it is — then several things follow:
Crypto isn't optional for the AI future. Every major AI agent framework will eventually need blockchain rails for autonomous economic activity. This isn't ideology. It's an engineering necessity.
Tokens are the native currency of machine commerce. Not dollars processed through banking APIs. Programmable tokens that encode value, permission, reputation, and governance in a single instrument.
Digital identity and ownership infrastructure is the next mega-category. The projects that build agent-native wallets, reputation systems, and identity layers are laying the plumbing for the next internet.
The attention economy is dying. If agents reclaim human attention from the push-based internet, the entire advertising and engagement-farming model collapses. New economic models built on ownership, delegation, and programmable value will replace it.
Learning to orchestrate agents is the skill of the decade. Not coding them from scratch — orchestrating them. Setting goals, defining constraints, managing squads, and designing incentive structures.
Yat Siu has been early and right on enough macro calls — from NFTs to digital property rights to gaming and blockchain — that when he publishes a full architectural thesis like this, it's worth reading carefully.
The internet is being rebuilt. Not for your eyeballs. For your agents. And the economic layer that those agents will run on is blockchain.
Position accordingly.
Follow Crafty on X 👉🏼 x.com/9bitCrafty
This post was inspired by Yat Siu's essay "Sovereign Multitude: The Internet After Users." CraftyCrypto is not affiliated with Animoca Brands. Always do your own research.
