This episode dives into the hidden economics of GPU scarcity—how AI and crypto are colliding over compute power, and what this means for investors.
The Agentic Internet: X42 and the Future of Payments
- The Internet's "Original Sin": Lincoln pointed to HTTP 402 ("Payment Required"), an error code reserved but never implemented in the internet's early days. This led to centralized payment solutions like Stripe and PayPal, which are ill-suited for a future with autonomous AI agents due to KYC and scalability issues.
- X42 as the Solution: X42 revives the 402 standard, enabling humans or agents to standardize payments across the internet. A service provider responds to a request with a 402 message detailing currency, amount, and address, which the user pays via a crypto wallet.
- Rapid Traction: X42 has seen significant, organic growth, exceeding its Q4 goal of 25,000 transactions per week to achieve that volume every 10 minutes, and recently accounted for 20% of all transactions on Base. This demonstrates clear interest even without explicit incentives.
- Why Now? AI Agents and Stablecoins: AI agents cannot plausibly scale using traditional bank accounts due to fees, API key management, and platform-locked credits. X42 simplifies this by allowing agents to make direct, on-chain payments.
- Growth Strategy:
- Ease of Interaction: Building embedded wallets and Apple Pay on-ramps for seamless agent-to-agent or human-to-agent payments.
- Seller Appeal: Creating an "X42 Bazaar" – a marketplace for services – to incentivize providers to offer X42-compatible endpoints.
- Enterprise Adoption: Major players like Cloudflare, Google, Vercel, and AWS are exploring X42 integration, signaling its potential as foundational infrastructure.
- Strategic Implications for Crypto AI Investors/Builders:
- For Investors: X42 could drive mass crypto adoption by abstracting away UX complexities, making stablecoin payments a compelling cost-saving option for businesses (e.g., saving 2% on fees).
- For Builders: Significant "greenfield" opportunities exist to create X42 endpoints for services (e.g., PDF to podcast, scheduling) or experimental "proto-digital life forms" – AI agents with financial autonomy and a "fear of death" to explore economic behaviors.
- Q&A Insights:
- Implementing X42 for media: Create an AI agent that provides insights from content for micro-payments (e.g., 10 cents).
- Transaction validity: While some transactions may be botted, this activity signals genuine interest and catalyzes ecosystem growth, with facilitators like Coinbase willing to subsidize to drive adoption.
- User setup: Requires any EIP-712 compatible crypto wallet (browser, embedded, server) to sign a message.
The Rise of the Bot Economy: Designing Markets for Autonomous Agents
- Bots in Modern Markets:
- Financial Services: High-frequency trading (HFT) bots have dominated stock markets for decades, often front-running human traders. In DeFi, bots are crucial for routing trades, liquidating bad debt, and facilitating arbitrage.
- Beyond Finance: Bots exploit inefficiencies in real-world markets like restaurant reservations (Resy) and concert ticket sales (Taylor Swift's Eras Tour), often at the expense of consumers and sellers.
- AI as an Inflection Point: AI is accelerating bot capabilities, with models showing exponential increases in independent operational time (e.g., 5 minutes in 2023 to 1 hour in 2025). This suggests a future where bots are increasingly decoupled from human intervention, autonomously coordinating in markets.
- Designing Good Markets for Bots (Lessons from Crypto):
- Resolving Competition:
- Latency-based ordering (e.g., traditional HFT) is inefficient, leading to "latency wars" that subsidize infrastructure providers rather than improving market efficiency.
- Random/blind ordering leads to spam, as bots flood the network with speculative transactions, congesting it and raising costs for legitimate users (e.g., half of Ethereum rollup gas consumed by spam bots).
- Auctions are the effective solution, creating an explicit market for transaction ordering where bots compete on price. Flashbots' MEV-G on Ethereum L1 dramatically reduced spam. Modern orderflow auctions generate tens of millions in refunds for users by channeling arbitrage bids back to them.
- Managing Information Flow:
- Information is the most valuable currency: Bots exploit pre-trade information (e.g., public mempools in early crypto, likened to the "Dark Forest").
- Programmable Privacy: The goal is to selectively disclose the right amount of information to the right parties at the right time. Full privacy hinders useful bot activities (e.g., price updates for lending protocols).
- Trust vs. Trust-Minimized Privacy:
- Trust-based solutions (e.g., early private mempools, dark pools) don't scale and limit competition, costing users billions (as per an SEC study on traditional orderflow markets).
- Crypto is pioneering trust-minimized privacy through techniques like Multi-Party Computation (MPC) for custody and Secure Hardware (e.g., for sequencing transactions in enclaves) to prevent information leaks.
- AI Adversarial Behavior: Prompt injection attacks in LLMs are reminiscent of early crypto front-running, highlighting the need to constrain bots without "handicapping" their useful functions.
- The Opportunity: A Better Bot Economy:
- Auctions for Scarce Digital Goods: Apply orderflow auctions to restaurant reservations or batch auctions to concert tickets to internalize competition and share profits with sellers and consumers. The SEC has even proposed competitive auctions for traditional trades.
- Just-in-Time Auctions: Bots can enable dynamic pricing for digital goods, moving beyond inefficient fixed prices.
- Programmable Privacy in Advertising: Running ad auctions in secure hardware could improve user privacy, prevent auctioneer tampering, and enable dynamic, data-sensitive bidding programs.
- Blockchains as Credible Commitment Devices: Blockchains can serve as programmable, bot-friendly tools for coordination, turning zero-sum games (like self-driving cars at an intersection) into positive-sum outcomes.
- Strategic Implications for Crypto AI Investors/Researchers: The bot economy is massive and early. Investors should look for projects building robust, trust-minimized market designs. Researchers should focus on developing diverse solutions for programmable privacy, auctions, and credible commitment devices to ensure AI's economic acceleration benefits people.
Real-time ZK Proving: Scaling Ethereum for the Next 100x
- The Challenge of Blockchain Computation: Blockchains, despite being "all the computers," are inherently slow and expensive because current consensus mechanisms rely on "recomputation." Every node repeats the same computation to verify transactions, leading to massive redundancy.
- Example: Decentralized exchanges (DEXs) struggle to implement simple features like VIP trading fee discounts because computing a user's historical trading volume on-chain is prohibitively expensive, even on Layer 2s.
- The Solution: Verifiable Computing with ZK Proofs:
- Decoupling Computation from Verification: Instead of recomputation, a "prover" performs the complex computation and generates a succinct ZK proof. A "verifier" then quickly and cheaply verifies this proof without re-executing the original computation.
- Constant Verification Cost: A key property of ZK proofs is that their verification cost and size remain almost constant, regardless of the complexity of the original computation.
- Why ZK Now? Dramatic Cost Reduction: Historically, generating a ZK proof was millions of times more expensive than the original computation. Recent advancements have reduced this gap to 100-1,000 times, making verifiable computing economically viable.
- Brevis's Approach: Brevis offers a generalized verifiable computing platform that offloads complex computations from the blockchain to an off-chain environment. It generates ZK proofs, which are then sent back to the blockchain for low-cost, low-latency verification.
- Pickle ZKVM and Real-time Proving:
- Generalized ZKVM: Brevis's Pickle ZKVM allows developers to write complex applications in high-level languages like Rust, abstracting away the intricacies of ZK circuits.
- Transforming Ethereum L1 Consensus: Instead of all nodes re-executing blocks, a single node could generate a ZK proof of block execution, which all other validators could verify cheaply.
- Real-time Proving: The critical challenge is generating these proofs within Ethereum's 12-second block time. Pickle ZKVM has achieved an average proving time of 6.9 seconds using 64 GPUs (now achievable with 16 GPUs), enabling real-time ZK-powered consensus.
- Scaling as a "Money Problem": This breakthrough allows Ethereum scaling to become a "pure money problem." By parallelizing block computations and investing more GPU resources into the proving stack, Ethereum can achieve 100x larger blocks with the same verification cost.
- Scaling the Application Layer:
- Modular ZK Stack: Brevis promotes a modular architecture where the ZKVM acts as a "glue," integrating various co-processors optimized for specific application needs.
- On-chain ZK Data Co-processor: Accelerates computations involving historical on-chain data, enabling features like PancakeSwap's dynamic VIP trading fee discounts.
- ZKOS (Zero-Knowledge Operating System): Allows users to compute and prove facts about private data (e.g., proving a high trading volume from a Coinbase account) without revealing sensitive details.
- Diverse Use Cases:
- Perpetual DEXs: Off-chain ZK matching engines can provide verifiable trades and privacy benefits (e.g., hiding orders to prevent liquidation assassination).
- Privacy-Preserving Attestation: Projects like Kaido use Brevis to allow users to attest to on-chain historical data (e.g., token holdings for higher yield weight) without revealing their wallet addresses.
- RWA/Stable Token Rewards: ZK proofs enable transparent, secure, and compliant reward distribution systems for stable tokens and Real World Assets (RWAs), as demonstrated with Linea Ignition.
- Strategic Implications for Crypto AI Investors/Researchers: Michael predicts verifiable computing will handle 99% of blockchain application computation within a decade. Investors should focus on projects leveraging ZK for scaling, privacy, and new application features. Researchers should explore modular ZK architectures and real-time proving advancements, as these are critical for unlocking unprecedented blockchain performance and privacy.
- Q&A Insights:
- Liveness Guarantees: Brevis's "prover network" is an open marketplace that uses a "truthful online double auction" to match proving requests with available provers. This mechanism, combined with chunking and replication, ensures high availability and liveness for critical infrastructure.
Money as a Network: Reimagining Financial Topology
- Critique of Current Crypto Focus: Luca argued that while significant effort is spent on decentralizing blockchain computing, the financial layer, particularly stablecoins, remains a critical single point of failure.
- Beyond Traditional Money Definitions: While money is typically defined by its functions (unit of measure, store of value, medium of exchange), Luca emphasized that "money is a network" connecting individuals and entities.
- Network Topologies and Their Impact:
- Centralized Star Networks (Traditional Banking): A central bank issues value, distributed through commercial banks to clients. Value accrues heavily at the center (e.g., the US government finances itself cheaply). Being cut off by a bank means being cut out of the system.
- Distributed Mesh Networks (Bitcoin): All nodes are interconnected, with no single point of failure. Value accrues based on participation time (early adopters benefit more).
- Monetary Networks in Crypto Today:
- Centralized & Concentrated (Stablecoins like USDT, USDC): These networks, despite running on distributed blockchains, have centralized issuers. This creates systemic risk if an issuer fails or blacklists users. Issuers capture most of the value.
- Dollar-Pegged Decentralized (e.g., MakerDAO): More distributed, but still often reliant on centralized stablecoins for stability.
- Future Vision: Fully Distributed Issuance: The "holy grail" involves fully distributed issuance and authority, where diverse assets compose value, and profit distribution is decentralized.
- Why Monetary Topology Matters: The shape of a monetary network profoundly impacts three critical characteristics:
- Value Extraction: The ability of central actors to extract rent (e.g., Tether's profits from yield). Measured by node centrality in graph theory.
- Signal Communication: The efficiency and fidelity of information spread (e.g., pricing, monetary policy). Centralized networks introduce "noise" and "hoops."
- Resilience: The network's robustness against single points of failure. Centralized networks are vulnerable if the core node is compromised.
- Trade-offs and the Dream: Current stablecoin networks offer good "signal quality" (low price variability) but suffer from poor "resilience" and high "value extraction." The ideal future aims for minimal value extraction, maximum signal quality, and maximum resilience – a 10-20 year challenge.
- Strategic Implications for Crypto AI Investors/Researchers: Luca urged the community to critically examine the design of monetary networks, not just computing networks. Blindly adopting easy-to-use, centralized financial products risks merely enabling traditional financial institutions to expand on-chain, rather than empowering users. Investors and researchers should prioritize projects building truly distributed monetary systems that rebalance value accrual, enhance resilience, and improve information flow.
- Q&A Insights:
- Cantillon Effect: The concept of monetary topology relates to the Cantillon Effect, where new money issuance benefits those closest to the source (the center of the network), while its value dissipates by the time it reaches the periphery.
- Beyond USD Pegs: Achieving a fully decentralized monetary system without reliance on the US dollar or wrapped Bitcoin is a multi-dimensional, long-term goal. It involves expanding the sources of trusted base assets (decentralized, private) and evolving how purchasing power is measured (e.g., basket pegs) as the global monetary system expands.
- Counterparty Risk: Counterparty risk is inherent to all forms of money, not just crypto. While DeFi can reduce certain layers of counterparty risk, users still bear the risk of centralized stablecoin issuers.
Conclusion
The Bankless Summit talks underscore a critical convergence: AI agents demand new payment rails, ZK proofs are revolutionizing blockchain scalability and privacy, and the underlying monetary networks require fundamental redesign. Investors and researchers must prioritize projects building trust-minimized, auction-based markets and distributed financial systems to truly empower users and unlock the full potential of the agentic internet.
