Circle launches Arc network to settle AI micro-payments
- Circle launched Arc, a blockchain using USDC gas to settle AI micro-transactions.
- Bitcoin developers are wiring Lightning wallets into AI models to pay for peer compute.
- Autonomous AI agents require permissionless digital rails because they cannot pass corporate compliance.
AI agents cannot open bank accounts or pass corporate compliance checks. They need native digital balance sheets.
On September 16, 2026, Circle Chief Technology Officer Nikhil Chandhok announced the mainnet launch of Arc on Bankless. The layer-one network replaces traditional gas tokens with native USDC, removing accounting hurdles for corporate treasuries. Using the Malachite consensus layer, Arc achieves payment finality in half a second. Chandhok emphasized that while the validator set remains permissioned for compliance, smart contract deployment is completely open. Crucially, Circle committed to zero chain rollbacks, choosing immutability over emergency bailouts during exploits.
"Infrastructure built for humans will not scale for machines."
- Nikhil Chandhok, Bankless
The network provides dedicated primitives built specifically for autonomous software. Arc includes an agent marketplace, reputation scoring, and Circle Nano payments designed to clear sub-cent transactions. Chandhok expects these micro-payment loops to let agents secure short-term loans, execute automated arbitrage strategies, and settle debt within seconds.
Two days later on Presidio Bitcoin Jam, host Max argued that Bitcoin offers a complementary, permissionless rail for machine-to-machine commerce. While institutional L1s rely on permissioned validators, AI agents buying raw compute need rails free from legacy KYC bottlenecks. Developers at Spiral are connecting Bitcoin directly into AI inference through Mesh LLM, allowing users to sell spare GPU cycles to developers in exchange for Lightning payments.
"AI agents cannot open bank accounts or submit tax forms. They need permissionless money that moves in milliseconds."
- Max, Presidio Bitcoin Jam
To demonstrate agent autonomy, Max configured GrokBot with a self-custodial Lightning wallet via Lexy. Operating inside an isolated cloud environment, the agent purchased an email address, registered domains, and published content on Stacker News to earn Bitcoin zaps - all without human code intervention. Presidio Bitcoin Jam co-host Steve cautioned that executing local agent software introduces severe security vulnerabilities if granted access to personal credential vaults, making isolated cloud virtual machines essential.
The broader landscape is shifting toward machine-first web architecture. Web assets are increasingly structured as LLM-readable text files - a practice host Max dubbed Agentic Engine Optimization - so searching agents can parse and execute transactions seamlessly. As fears of autonomous agent swarms grow among regulators, decentralized peer-to-peer compute networks provide a resilient fallback against prospective internet shutdowns.
The web is being rebuilt for silicon consumers. The battle is over who controls their wallet.
Source Intelligence
- Deep dive into what was said in the episodes
Grokbot and Muse Take Agents Mainstream, Bitcoin’s Agent Moment, Dario vs Jensen on AI Pacing • Sep 18
- Max highlights Muse Spark and GrokBot's ability to instantly spin up virtual private servers in the cloud to isolate active agents. Steve cautions that installing any native software on a personal laptop introduces severe local attack vectors.
- Max configured GrokBot to run a self-custodial Lightning wallet through Lexy, allowing the agent to purchase an email address and transact autonomously. The agent successfully navigated micro-payments to purchase domain names and post content on Stacker News.
- Max asserts that the next phase of Bitcoin will be driven by autonomous agents requiring frictionless, millisecond-level identity and wallet setup. Agents must operate on permissionless rails because they cannot satisfy traditional legacy KYC frameworks.
- Max outlines Agentic Engine Optimization, where developers structure web assets like LLM text files so autonomous searching agents can parse them. Traditional search optimization is shifting toward catering directly to agent workflows rather than human users.
Also discussed on this episode: (7)
Agents (1)
- Max transitioned his weekly preparation workflow from Signal to Buzz to utilize AI agents that can read X links directly. Max installed the GrokBot command-line interface on Buzz to generate automated weekly summaries.
Big Tech (1)
- Max argues Elon Musk's purchase of X will prove to be a massive bargain. The acquisition secures both invaluable training data and unparalleled distribution for AI development.
AI Infrastructure (1)
- Steve and DK explain Spiral's Mesh LLM project, which creates a peer-to-peer network allowing users to monetize spare compute cycles on local machines. Developers can buy model inference directly from the mesh using permissionless Bitcoin payments.
Regulation (1)
- DK reports that the Blockchain Regulatory Clarity Act failed to pass after being significantly weakened in its final hours. The bill intended to exempt lightning service providers and developers from burdensome money services business regulations.
Safety (2)
- Max suggests governments should monitor the physical supply chain of hazardous raw materials instead of trying to regulate model weights. Restricting chemical and biological ingredients is far more enforceable than controlling digital information.
- DK argues that the mid-2024 OpenAI security breach proves closed-weight AI repositories function as dangerous centralized honeypots. Restricting model access to a few large corporations increases systemic vulnerability compared to decentralized, open-weight alternatives.
Open Source (1)
- DK reveals Spiral's plans to produce a puppet-based film advocating for open-source AI and open-weight models. The project builds on the team's prior niche software release, Puppet Jack, to target a much broader technology audience.
Arc Mainnet, AI Agents, and Tokenized Markets | Nikhil Chandhok, CTO of Circle • Sep 16
- ARK achieves payment finality in half a second using the Malachite consensus layer. Nikhil Chandhok explains this rapid settlement minimizes transaction risk, which is essential for high-frequency financial applications and automated agents.
- Circle is building lower-level primitives for AI agents on ARK, including an agent marketplace, reputation systems, and micro-payment rails. Nikhil Chandhok expects these tools to turn autonomous agents into fully functional economic actors capable of securing loans.
Also discussed on this episode: (8)
Stablecoins (4)
- Circle launched its mainnet, ARK, which Nikhil Chandhok describes as an economic operating system. The platform aims to solve full-stack problems for developers, users, and financial institutions by providing native primitives from the network layer up to applications.
- ARK eliminates the need for native network tokens by allowing users to pay gas fees in stablecoins like USDC. Nikhil Chandhok claims this removes a major accounting and treasury management hurdle for traditional corporations transacting in real-world assets.
- Nikhil Chandhok anticipates that global stablecoin laws will evolve rapidly once US legislation takes effect. Rather than issuing local stablecoins in 190 countries, Circle plans to partner with regional issuers and connect them to ARK-native liquidity pools.
- Circle has processed nearly one trillion dollars in off-ramps over the lifetime of USDC. This massive bridge between traditional bank dollars and digital currency positions the firm as a primary gateway for institutional on-chain activity.
Enterprise (1)
- ARK features a permissioned validator set to build trust for institutional traffic, but maintains permissionless smart contract deployment. Developers do not need Circle sign-up or specialized developer keys to launch applications on the network.
Privacy (1)
- ARK implements privacy through trusted execution environments, allowing users to toggle between public and private transactions in a single wallet. Nikhil Chandhok states that neither validators nor Circle can access the contents of these shielded transactions.
Protocol (2)
- ARK is a fully immutable blockchain with over 20 validators. Nikhil Chandhok asserts that Circle will not roll back transactions or freeze funds to reverse exploits, as maintaining strict immutability is critical to preserving trust in public infrastructure.
- Circle spent close to 18 months developing the ARK L1 project prior to mainnet launch. The near-term roadmap includes releasing advanced privacy features, adding more validators, and eventually transitioning to a proof of stake consensus mechanism.