Nexus

The first wave of onchain builders came from cypherpunk and finance backgrounds. After talking to guests on the Exponential podcast over the past few months, I've realized the next wave is starting to look different. Today's builders are AI engineers working on agent infrastructure. Or community leaders building developer pipelines for underrepresented groups. Maggie Love from SheFi joined me this week to talk about exactly that shift, what it takes to build a developer community around agentic finance, and what the onboarding path looks like when your audience isn't already fluent in crypto-native tooling. If you're thinking about who actually builds the next decade of this industry, it's worth a listen.

Most discussions about AI agents frame them as tools — faster software, better automation, incremental productivity. The more interesting framing, from an organizataional perspective, is that an AI agent increasingly looks like a permanent team member. Not a subscription software license. A tool is something you use. A teammate is something you build around. For engineering, finance, and operations teams thinking about where to deploy agents next, this distinction is non-trivial. It changes what you measure, how you onboard, and what infrastructure the agent actually needs to function reliably. Nexus and Node Ops on Exponential talking about what that looks like in practice.

One billion transactions per second. That's not a roadmap milestone. It's the baseline for payment infrastructure built for the internet. The ceiling on most blockchains — even the fast ones — isn't raw processing speed. It's verification architecture. A system that replays every computation across every validator can't reach a billion TPS. The overhead compounds with each transaction. Proof-based execution changes the math. Execute once, generate a cryptographic proof that the computation was correct, and let every validator verify that proof. The cost stays constant, no matter how complex the computation.

Every blockchain built in the last decade made the same implicit assumption: the user is a human. That assumption is becoming false — fast. Stripe's founders said it directly in their 2025 annual letter: AI agents will likely be responsible for most internet transactions, and the infrastructure will need to support more than one billion transactions per second to keep up. The agents aren't coming. They're here. Solana has processed 15 million onchain agent transactions. AI now drives an estimated 65% of crypto trading volume. The user base is shifting under the feet of protocols still designed for the other 35%. The problem: everything about how blockchains were built assumes human latency tolerance. Ethereum's 12-second block time, dynamic gas pricing, browser-extension onboarding, RPC rate limits designed for occasional requests. These were reasonable choices when the user was a person at a screen. A trading agent running thousands of transactions per hour can't absorb a random gas spike mid-execution. It can't navigate a browser wallet. And critically — it can't look at a confirmation screen and make a judgment call. It needs cryptographic proof that its counterparts executed correctly. That last requirement is the one most people miss. In multi-agent systems, "Agent B said it worked" isn't good enough. You need "Agent B can prove it worked." That's a different infrastructure problem entirely. At Nexus, we've been building for this from the start: sub-200ms finality, API-first access, CCXT-compatible trading interfaces, and a zkVM that lets agents verify each other's execution cryptographically. Full post in the comments.

In the latest Exponential, we talked with Mrudul Gole, Head of Business Development at NodeOps, about how a company that started running blockchain validators ended up building an AI deployment platform — and what that journey reveals about where compute is headed. A few of the highlights from the episode: The performance gap between frontier models and open-source models is way narrower than most teams assume. The cost gap, though, is massive. Teams that actually test this end up cutting model costs significantly — and latency ends up mattering more than benchmarks in production. On the blockchain/AI intersection: major exchanges are now exposing APIs through MCPs, opening the door to AI-driven trading in a way that hasn't existed before. Agentic payments via stablecoins are another area worth watching closely. A bold prediction: within the next six months, most small dev teams will have at least one AI agent as a permanent member of the team — not a tool you open occasionally, but something that's watching your infrastructure, handling deployments, and escalating to humans only when real judgment is needed. This episode is worth a listen if you're building at the intersection of AI and blockchain, or just thinking hard about what the agentic era means for how we work. The show link is in the comments.

AI agents have already executed hundreds of millions of financial transactions in 2026. They're trading, routing stablecoin payments, and managing positions… autonomously, without a human in the loop. The infrastructure they're running on was not built for them. Most chains record that transactions happened. None of them can prove the computation was correct. That distinction is academic when a human trader is watching their own positions. It becomes commercially and legally significant when an autonomous agent is the one making decisions. Messari just published technical breakdown of Nexus. Here's their framing of what we are building: "High-performance financial infrastructure should not be simulated through smart contracts, but embedded directly into the base layer and verified cryptographically." Every financial computation — order matching, margin calculation, liquidation — generates a cryptographic proof. An AI agent running on Nexus won't just act, it will leave a receipt that anyone can verify. The Nexus Exchange will be embedded directly into the L1 as a native co-processor. USDX will be the native settlement asset. Agents inherit the full financial stack and every operation will be provably correct. Find the link to the full report in the comments.

What if the biggest win for crypto is the moment nobody notices it? That's the vision Nic Roberts-Huntley, founder and CEO of Blueprint Finance, laid out on the latest episode of Exponential. His firm is building institutional-grade DeFi yield infrastructure — vaults that take your assets, keep your exposure intact, and quietly put your capital to work across blue-chip DeFi protocols. No late nights in Discord required. But the longer game is even more interesting. Roberts-Huntley believes institutions may leapfrog the slow retail-first adoption curve and move directly onchain — and that vaults become the connective tissue absorbing those flows. His 24-month goal: everyday people receiving DeFi-generated value inside their savings accounts, credit cards, and 401ks without ever knowing it happened. "My most exciting day will be knowing that there is somebody who has no idea that they just received value within their savings account from a DeFi-type mechanism." The technology wins by disappearing. Full conversation on Exponential — link in comments.

Nexus founder and CEO Daniel Marin joined the Exponential podcast this week to give an update about the company's plans, make the case that AI is still underestimated, and explain why he thinks financial infrastructure for autonomous agents is the biggest opportunity in crypto right now. https://lnkd.in/gRBv5-G5

Notes on sustainable economics and protocol design.

Laszlo S., co-founder and CEO of Kiln, joins Exponential to unpack one of the most important trends in crypto right now: the evolution of yield. The conversation explores how staking, onchain lending, and tokenized real-world assets are reshaping how digital assets generate returns, and why this shift could transform the relationship between DeFi and traditional finance. Laszlo also explains Kiln’s new RailNet infrastructure and shares why he believes a large portion of global capital markets will move onchain in the next decade. Full episode: https://lnkd.in/gZfTWz7y