𝐆𝐞𝐫𝐚𝐥𝐝𝐂𝖗𝐲𝐩𝖙 人

We all thought @arc had been forgotten, but its TVL has been quietly growing while everyone was looking elsewhere. I went through DefiLlama's data, and @arc is now sitting at roughly $633M in DeFi TVL. What's even more surprising is where that money is going: • @Morpho Blue: $298.5M • @aave V4: $184.7M • Extended: $106.3M That's nearly $483M sitting in Morpho and Aave alone, accounting for about 77% of Arc's TVL. And lending isn't the only activity picking up. Arc also has around $1.16B in stablecoin supply, with weekly perpetual trading volume approaching $1.9B. But I wouldn't rush to call this organic growth just yet. Depositing capital into lending protocols is one thing. Having people actively borrow against it is another. I'd be watching: • how much of that liquidity is actually being borrowed • whether trading volume holds up • how much capital stays when incentives slow down @arc may have gone quiet on the timeline, but the numbers suggest people haven't stopped putting capital to work there.

An AI agent wants to buy $0.10 worth of market data. The data costs 10 cents, but processing the payment through a conventional card could cost another 30 cents. Imagine that same agent making thousands of small purchases every day. Suddenly, the cost of moving money becomes a bigger problem than the service itself. This is one reason I've been paying closer attention to the 𝗲𝗺𝗲𝗿𝗴𝗶𝗻𝗴 𝗽𝗮𝘆𝗺𝗲𝗻𝘁 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 𝗳𝗼𝗿 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀. And the numbers are getting difficult to ignore. According to @chainalysis, x402 recorded approximately 138.5 million cumulative transactions on Base through Q3 2026. Over 87% of the payments analyzed were worth $0.10 or less. That's a fascinating payment pattern because AI agents don't necessarily spend like humans. An autonomous research agent might need to purchase market data, pay for inference, access an API, and run security checks, all within a single workflow. Charging a conventional checkout fee for every tiny purchase becomes economically inefficient. 𝗧𝗵𝗶𝘀 𝗶𝘀 𝘄𝗵𝗲𝗿𝗲 𝘅𝟰𝟬𝟮 𝗲𝗻𝘁𝗲𝗿𝘀 𝘁𝗵𝗲 𝗽𝗶𝗰𝘁𝘂𝗿𝗲. Developed by @CoinbaseDev, x402 allows software to: → request a resource → receive payment instructions through HTTP 402 → authorize a payment → access the service after verification Coinbase's facilitator offers 1,000 free monthly settlements, followed by $0.001 per settlement under its published pricing. But payment settlement is only one part of this emerging market. I've been following several companies building different pieces of the infrastructure: • @CoinbaseDev — x402 payment infrastructure for machine-to-machine transactions. • @crossmint — Agent wallets, stablecoins, existing card integration, and merchant checkout. • @turnkeyhq — Wallet infrastructure that developers can integrate into agent applications. • @evervault and @basistheory — Secure payment credentials and sensitive-data infrastructure. • @nevermined_ai — Usage-based payments, metering, and x402 settlement for AI services. • @Stripe — Shared Payment Tokens that allow agents to initiate authorized purchases without exposing underlying card credentials. • @Visa — Intelligent Commerce and Trusted Agent Protocol for recognizing and authorizing agent-driven transactions. • @Mastercard — Agent Pay for Machines, combining agent credentials, spending permissions, and multiple settlement rails. Notice something here? 𝗧𝗵𝗲 𝗰𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻 𝗶𝘀𝗻'𝘁 𝘀𝗶𝗺𝗽𝗹𝘆 𝗰𝗿𝘆𝗽𝘁𝗼 𝘃𝗲𝗿𝘀𝘂𝘀 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗽𝗮𝘆𝗺𝗲𝗻𝘁𝘀. It’s about which infrastructure can give machines purchasing power without sacrificing cost efficiency, security, or accountability. And there’s still an uncomfortable question: If an AI agent makes an unauthorized purchase, exceeds its spending limit, or pays for a service that never arrives, who takes responsibility? Cheap settlement doesn’t automatically solve those problems. There’s also a reason to be cautious about the adoption numbers. Chainalysis identified speculative activity contributing to earlier x402 growth, so transaction counts alone don’t prove sustainable commercial demand. My expectation is that agent payments will develop across multiple rails: ▸ Stablecoins and x402 have compelling economics for micropayments. ▸ Card networks retain established merchant acceptance and dispute mechanisms. The infrastructure that connects both while enforcing verifiable spending permissions, could become an important part of autonomous commerce. AI agents are learning to do work independently. The next challenge is allowing them to pay for that work without losing human control over the money.