The Intersection of AI and Web3 (Part 2)

Over a year ago, I wrote about the convergence of AI and Web3, arguing that we were witnessing the beginning of an unprecedented transformation. At the time, we were exploring possibilities—decentralized computing, identity authentication, blockchain optimization with AI, and real-time data streams powering intelligent systems. Today, those possibilities have crystallized into production systems, regulatory frameworks, and a thriving ecosystem of decentralized AI infrastructure.

As a full‑stack engineer building at this intersection, I'm seeing the narrative shift from "what if" to "what now." The infrastructure is maturing. The regulatory uncertainty is clearing. And most importantly, the use cases are real—AI agents transacting autonomously, decentralized marketplaces for AI services, individuals monetizing their data, and cross-chain intent-based systems coordinating complex financial operations.

This is the update I wish I'd had when I started building in this space.

AI Web3

The GENIUS Act: Why Regulatory Clarity Matters for AI × Web3

One of the most significant developments in 2025 has been the enactment of the GENIUS Act (Guiding and Establishing National Innovation for U.S. Stablecoins Act), signed into law by President Donald Trump in July 2025. While ostensibly about stablecoins, this legislation has profound implications for the AI and Web3 convergence.

Here's why it matters:

Programmable money for AI agents: As I explored in my post on Agentic Finance, autonomous AI agents need stable, programmable rails to transact safely. The GENIUS Act legitimizes stablecoins as payment infrastructure, which means AI agents can now operate within a regulated framework—critical for institutional adoption.
Regulatory clarity enables innovation: For engineers building AI × Web3 systems, the biggest blocker has been regulatory uncertainty. With clear guidelines around reserve requirements ($1 of reserves per $1 of stablecoins issued), AML/KYC compliance, and issuer licensing, builders now have a stable foundation to work from.
Stablecoins as the settlement layer: Stablecoins have become the de facto standard for AI agent transactions. Protocols like Google's AP2 and Coinbase's x402 enable AI agents to use stablecoins under cryptographic constraints. With regulatory legitimacy, this layer can scale beyond crypto-native users into mainstream fintech.

For a deeper dive into how stablecoins are becoming financial infrastructure, see my post on Stablecoin Summer. The key insight: stablecoins are the connective tissue between AI agents and real-world value transfer.

Agentic Finance: AI Agents as Economic Actors

In my Agentic Finance post, I argued that we're entering a paradigm where AI agents don't just advise—they transact, execute, and move capital autonomously. Since then, the infrastructure for agentic finance has matured significantly.

What's changed:

Standardized protocols for agent payments: Google's AP2 protocol and Coinbase's x402 implementation provide the foundational standards for secure multi-party computation in finance. AI agents can now delegate intent, verify execution, and maintain cryptographic guarantees without compromising user sovereignty.
Non-custodial agent wallets: Platforms like Crossmint's Agentic Finance provide AI agents with non-custodial wallets equipped with programmable guardrails. This means agents can transact within user-defined boundaries—spending limits, approved contracts, time-based restrictions—without requiring constant human approval.
Cross-chain agent coordination: The convergence of intent-based DeFi architecture and AI agents is particularly powerful. Agents can now express financial intents ("swap 1 ETH for at least 3,000 USDC on the cheapest chain") and solver networks fulfill them—abstracting away multichain complexity.

Real-world adoption:

According to a 2025 Wolters Kluwer survey, while only 6% of finance leaders currently employ agentic AI, an additional 38% plan to adopt it within the next year—a projected sixfold increase. However, Gartner predicts that over 40% of agentic AI projects may be discontinued by 2027 due to escalating costs and unclear business value. This underscores the importance of strategic implementation and clear ROI when integrating agentic AI into business processes.

For builders: agentic finance is production-ready, but it requires thoughtful design. Agents need deterministic, constraint-based execution models to operate safely. Intents provide exactly that.

Decentralized AI Marketplaces: The New Primitives

One of the most exciting developments is the emergence of decentralized AI marketplaces—platforms where AI models, datasets, and computational resources are bought and sold without intermediaries. These marketplaces are redefining how AI services are accessed, monetized, and governed.

The Core Value Propositions

1. Data Ownership and Monetization

In traditional AI systems, large corporations monopolize data. Decentralized AI marketplaces flip this model: individuals retain ownership and control over their data, choosing to keep it private or monetize it through blockchain-based marketplaces.

Platforms like BitSeek are pioneering models where users can:

Contribute training data and receive compensation in tokens
Maintain privacy through cryptographic proofs (e.g., zero-knowledge proofs)
Participate in decentralized data collectives that aggregate and anonymize data

2. Specialized AI Services on Marketplaces

Instead of relying on centralized AI providers (OpenAI, Anthropic, Google), developers can now access specialized AI models through decentralized marketplaces. This enables:

Model diversity: Access to niche models fine-tuned for specific domains (legal, medical, financial)
Price competition: Market-driven pricing rather than monopolistic pricing
Composability: Mixing and matching models for complex workflows

3. Decentralized Computing Power

Training and running AI models requires significant compute. Decentralized networks enable individuals to monetize unused computing power by contributing it to AI workloads.

Leading Decentralized AI Marketplaces

Here's a snapshot of the most significant players in this space:

Bittensor (TAO): The Neural Internet

Bittensor operates as a decentralized network with specialized subnets for various AI tasks. It creates a competitive environment where only the most efficient models are rewarded through a tokenized incentive mechanism.

Architecture: Peer-to-peer network of AI models competing on performance and efficiency
Use cases: Text generation, image synthesis, prediction markets, data retrieval
Key innovation: Models are validated by the network, ensuring quality through economic incentives

Fetch.ai (FET): Autonomous Economic Agents

Fetch.ai has deployed over 23,000 autonomous economic agents capable of performing tasks ranging from route planning for delivery robots to smart energy management.

Architecture: Multi-agent systems with decentralized coordination
Use cases: Supply chain optimization, energy grid management, DeFi automation
Key innovation: Agents negotiate, transact, and learn autonomously within economic frameworks

SingularityNET (AGIX): The AI Services Marketplace

An open marketplace for AI services where developers monetize AI algorithms and consumers access a diverse array of AI tools within a decentralized framework.

Architecture: Marketplace with reputation systems and decentralized governance
Use cases: Natural language processing, computer vision, robotics, bioinformatics
Key innovation: Interoperability between AI services through standardized APIs

Ocean Protocol (OCEAN): Data Marketplaces

Ocean Protocol enables data providers to monetize datasets while maintaining control and privacy through blockchain-based access control and compute-to-data paradigms.

Architecture: Decentralized data exchange with privacy-preserving computation
Use cases: Medical data sharing, financial data analytics, IoT data monetization
Key innovation: Data remains with the provider; buyers run computations on encrypted data

Morpheus: Open-Source Personal AI

As I mentioned in my original AI and Web3 post, Morpheus incentivizes the first open-source peer-to-peer network of personal general-purpose AI. The MOR token rewards contributors who build, maintain, and utilize the open-source decentralized AI infrastructure.

Architecture: Decentralized smart agent network with tokenized incentives
Use cases: Personal AI assistants, automation, research, creative tools
Key innovation: Truly user-owned AI that isn't controlled by a single corporation

NodeGoAI: Monetizing Unused Compute

Founded in 2021, NodeGoAI enables users to monetize unused computing power by transforming it into resources for AI and high-performance computing applications.

Architecture: Peer-to-peer distributed computing network
Use cases: AI model training, rendering, scientific computation
Key innovation: Accessible AI compute for developers without access to expensive cloud resources

OORT: Decentralized Data Cloud for AI

Established in 2021, OORT offers a decentralized data cloud for AI, collaborating with major tech companies to provide data storage and processing solutions.

Architecture: Blockchain-based distributed storage with verification layers
Use cases: AI training data storage, model hosting, decentralized CDN
Key innovation: Cost-effective, censorship-resistant data infrastructure for AI systems

Data Ownership: The Economic Revolution

The shift from centralized to decentralized AI fundamentally reimagines data economics. In Web2, your data is extracted, aggregated, and monetized by platforms—without your consent or compensation. In Web3, you control your data and decide how it's used.

How it works in practice:

Data contribution and tokenization: You contribute data (e.g., training examples, sensor data, medical records) to a decentralized marketplace. Your contribution is recorded on-chain and tied to a unique identifier (wallet address).
Access control via smart contracts: Data access is governed by smart contracts. Buyers request access, and you approve (or reject) based on terms you define—price, usage restrictions, time limits.
Compensation in tokens: When your data is accessed or used to train models, you receive compensation in protocol tokens. This creates passive income streams for data contributors.
Privacy-preserving computation: Techniques like federated learning, homomorphic encryption, and zero-knowledge proofs enable AI models to train on your data without ever accessing the raw data itself.

Why this matters:

Fairness: Data creators are compensated for their contributions, addressing a longstanding ethical issue in AI development.
Incentive alignment: High-quality data is rewarded, creating incentives for accurate, diverse, and representative datasets.
Privacy: Users maintain control and can revoke access, a stark contrast to the "data is the new oil" extraction model.

For engineers: data marketplaces are becoming critical infrastructure. If you're building AI features, consider integrating decentralized data sources rather than scraping or purchasing from data brokers.

The Blooming Ecosystem: AI × Web3 Startups and Networks

Beyond the major platforms, a vibrant ecosystem of AI and Web3 startups is emerging. Here's a curated list of notable projects pushing the boundaries:

AI Infrastructure and Compute

Akash Network: Decentralized cloud compute marketplace for AI workloads
Render Network: Distributed GPU rendering and AI computation
Gensyn: Verifiable deep learning computation on decentralized networks
io.net: Decentralized GPU network for AI model training and inference

AI Agents and Automation

Autonolas: Protocol for building autonomous AI agents with economic coordination
Olas: Unified network of off-chain services for blockchain automation
ChainML: On-chain AI inference and model hosting
MyShell: Decentralized AI assistant marketplace

Data and Model Marketplaces

Grass: Decentralized web scraping network for AI training data
Synapse Protocol: Cross-chain AI model marketplace
Nous Research: Open-source AI models with decentralized governance
Hyperbolic: Verifiable AI inference marketplace

Web3 × AI Tooling

Nillion: Privacy-preserving computation network for AI
Ritual: AI coprocessor for blockchains—bringing AI inference on-chain
Oraichain: AI-powered oracle and data hub for smart contracts
Phala Network: Confidential AI computation using TEEs (Trusted Execution Environments)

Identity and Reputation

Worldcoin: Proof of personhood using biometric verification to distinguish humans from AI
Gitcoin Passport: Decentralized identity with humanity verification for AI-resistant sybil protection
BrightID: Social identity network preventing duplicate accounts in AI-dominated spaces

Emerging Projects to Watch

Vana: User-owned data treasury for AI training
Prime Intellect: Decentralized AI research collaboration
Allora Network: Decentralized machine intelligence network
Sahara: Decentralized AI blockchain with knowledge tokenization

This is just a snapshot—new projects emerge weekly. The key pattern: AI and Web3 are no longer separate ecosystems; they're converging into a unified infrastructure for intelligent, decentralized systems.

The Technical Convergence: Where It All Comes Together

The real power of AI × Web3 emerges when you combine the primitives I've been writing about:

Stablecoins as payment rails (Stablecoin Summer): AI agents transact using regulated, programmable money
Intent-based execution (Multichain DeFi Intents): AI agents express financial goals, solver networks fulfill them across chains
Agentic finance (Agentic Finance): AI agents operate autonomously within cryptographic guardrails
Decentralized AI marketplaces: AI models, data, and compute are accessible without centralized gatekeepers
Model Context Protocol (MCP): Standardized interfaces for AI models to interact with external tools and data

A concrete example:

Imagine an AI agent managing a DeFi portfolio:

The agent analyzes market conditions using decentralized data feeds (Ocean Protocol)
It identifies an arbitrage opportunity across chains
It expresses an intent: "Swap 10,000 USDC on Arbitrum for wBTC on Base, maximize output"
Solver networks (UniswapX, Across Protocol) compete to fulfill the intent
The agent executes the trade using a non-custodial wallet with programmatic guardrails
Settlement occurs in stablecoins (GENIUS Act-compliant rails)
The agent logs the transaction to shared memory (MCP-compliant architecture)

This isn't a prototype—it's how AI-driven financial systems are being built today.

Product and Engineering Implications

If you're building at the intersection of AI and Web3, here's what you should prioritize:

1. Design for Agent-First UX

Traditional UX design assumes human users clicking buttons. Agentic systems require API-first, intent-driven interfaces where agents express goals and the system fulfills them.

Key considerations:

Declarative APIs over imperative commands
Constraints and guardrails built into smart contracts
Cryptographic verification for agent actions

2. Leverage Decentralized AI Services

Instead of hardcoding OpenAI or Anthropic APIs, integrate with decentralized AI marketplaces. This provides:

Resilience: No single point of failure
Cost efficiency: Market-driven pricing
Specialization: Access to niche models optimized for your domain

3. Implement Data Sovereignty

If your product collects user data, give users control over it. Integrate with decentralized data marketplaces where users can monetize their contributions.

Technical patterns:

User-controlled data vaults (e.g., Vana, Ocean Protocol)
Privacy-preserving ML (federated learning, differential privacy)
Tokenized data contribution rewards

4. Build with Model Context Protocol (MCP)

MCP standardizes how AI models interact with external tools and data. As I explored in my MCP post, this creates modular, composable AI systems.

Key benefits:

Separation of model logic and business logic
Reusable tool definitions across models
Persistent memory and execution logs

5. Use Stablecoins and Intent-Based Rails

For any financial feature, default to:

Stablecoin settlement (GENIUS Act-compliant)
Intent-based execution (CoW Protocol, UniswapX, Across)
Cross-chain abstraction (Wormhole, Axelar)

This gives you instant settlement, MEV protection, and multichain reach without custom bridge integrations.

Challenges and Open Questions

I'm bullish—but pragmatic. Here are the hard problems we're still solving:

1. Scalability of Decentralized AI

Running AI models on-chain or in decentralized networks is expensive and slow. Most projects use hybrid architectures (off-chain compute, on-chain verification)—but verifiability remains a challenge.

Open questions:

How do we verify AI inference without re-running the computation?
Can zero-knowledge proofs make AI inference verifiable at scale?

2. Model Quality and Adversarial Risks

Decentralized AI marketplaces face the challenge of model quality assurance. Without centralized curation, how do you prevent malicious or poorly performing models?

Emerging solutions:

Reputation systems (staking, slashing)
Verifiable benchmarks on standardized datasets
Economic incentives (Bittensor's validation mechanism)

3. Regulatory Uncertainty Beyond Stablecoins

While the GENIUS Act provides clarity for stablecoins, most AI × Web3 use cases remain in regulatory gray zones. Questions around data privacy, model liability, and cross-border AI agent transactions are unresolved.

Build defensively:

Prioritize privacy-preserving techniques
Design for regulatory compliance (AML/KYC hooks)
Monitor jurisdictional developments

4. Composability vs. Fragmentation

The ecosystem is growing rapidly—but with dozens of competing protocols, standards, and networks. How do we achieve interoperability?

Promising directions:

Standardized agent protocols (AP2, x402)
Cross-chain messaging (Wormhole, Axelar, LayerZero)
Universal intent formats (ERC-7683)

5. Cost and Accessibility

Decentralized AI infrastructure is more expensive than centralized alternatives (for now). This creates adoption friction, especially for developers used to cheap API calls from OpenAI or Google.

Mitigation strategies:

Subsidized compute for early adopters (protocol incentives)
Hybrid models (centralized inference, decentralized training)
Economies of scale as networks mature

Why I'm Doubling Down

When I wrote The Intersection of AI and Web3 in 2024, I was exploring a thesis. Today, that thesis has hardened into conviction.

Here's why:

The infrastructure is real: Decentralized AI marketplaces are processing real transactions. AI agents are transacting with real money. Intent-based execution is handling billions in monthly volume.
Regulatory clarity is emerging: The GENIUS Act is just the beginning. As stablecoins and AI agents scale, more regulatory frameworks will follow—enabling institutional adoption.
The incentives align: Decentralized models align incentives between data providers, model developers, and consumers in ways centralized platforms never could.
The technical primitives are composable: Stablecoins + intents + AI agents + decentralized marketplaces = a unified infrastructure stack for intelligent, autonomous systems.
The problems being solved are real: Data ownership, model accessibility, compute costs, cross-chain coordination—these aren't hypothetical issues. They're blocking real products. Decentralized AI × Web3 infrastructure solves them.

What to Watch in 2026

Here's what I'm tracking as critical signals:

AI agent adoption in DeFi: Will intent-based protocols integrate AI agent execution as a standard feature?
Decentralized AI marketplace traction: Which platforms achieve product-market fit beyond crypto-native users?
Data monetization models: Will decentralized data marketplaces attract mainstream data contributors?
Agentic finance regulations: How will regulators respond to autonomous AI agents transacting at scale?
Cross-chain AI agent coordination: Will universal intent standards (ERC-7683) enable seamless AI agent interoperability?

The Convergence Accelerates

The intersection of AI and Web3 isn't a speculative future anymore—it's an emerging infrastructure layer that's reshaping how we build intelligent systems, manage data, and coordinate economic activity.

We've moved from exploring possibilities to operationalizing them. AI agents are transacting autonomously. Decentralized AI marketplaces are enabling data sovereignty and model accessibility. Stablecoins are providing programmable money rails. Intent-based execution is abstracting multichain complexity.

For engineers, entrepreneurs, and builders: the infrastructure is here. The regulatory environment is clarifying. The ecosystem is thriving. The use cases are real.

The question isn't whether AI and Web3 will converge—they already have. The question is: what are you building with it?

I'm continuing to explore this convergence through my work on modern Web3 architecture, intent-based DeFi, and agentic finance. If you're building in this space, I'd love to hear what you're working on.

Join me on this journey as we build the next generation of intelligent, decentralized systems—where AI and Web3 converge to create a more open, transparent, and equitable digital future.