Letters from the CEO

This blog post marks the first installment of a new quarterly series from the Mysten Labs leadership team, “Letters from the CEO,” in which you’ll hear the latest updates: what we’re working on, what we’re excited about, and what we’re looking forward to. Settle in – there’s a lot to cover. 

First, many of the team just returned from Sui Basecamp, co-located with TOKEN2049 Singapore, at the iconic Marina Bay Sands. The third iteration of Sui’s annual conference was nothing short of energizing – over 4500 attendees from all over the ecosystem gathered to showcase their work, build together, and connect. As a remote company, we treasure any chance to gather in person, but attending Sui Basecamp takes it a step further: we go from connecting with colleagues in real life to meeting our global ecosystem. Sui Basecamp is also the moment we get to unveil what we’ve been working on. 

Brave New World 

We are living through the world’s next massive technological transformation, and the direction of travel is clear: AI agents are taking on actions of increasing consequence. Agents now book, buy, pay, negotiate, decide, and ship code on our behalf. Personal agents like Meta’s Muse have rocketed to the top of App Stores, and people are handing them actions that matter: moving money, making commitments, touching sensitive data. The same is true for enterprise workflows. 40% of enterprise applications are projected to include task-specific AI agents by the end of 2026. This is up from less than 5% a year ago.

The problem, and our opportunity, is trust.

Today, we cannot trust that AI will act reliably, under human control, and not cause unintended harm. Part of this is structural: LLMs are stochastic, and their outputs are not fully deterministic. This may be fine for non-critical tasks, but incompatible with systems that demand defined, bounded, and provable outcomes. Part of the reason is systemic: no single party controls the entire system or is liable for damages when the unexpected happens. When something goes wrong, no credibly neutral 3rd party can be called upon to adjudicate.

The result is a ceiling on the usefulness of AI. Nobody hands more work, more money, or more responsibility to a system they can’t trust.

What We’re Building

Mysten Labs is now building the products and services that let people and businesses trust agents, and let agents trust each other. This will have two outcomes:

• Control: agents act within the bounds defined by their people, and nothing more.

• Enablement: agents can safely perform tasks autonomously without humans in the loop.

Everything we’re building will be anchored to Sui and Walrus. I invite you to watch my keynote from Sui Basecamp for a sneak peek. 

This new direction is born of a fundamentally optimistic belief in the power of AI to improve our lives and advance humanity’s goals. From transacting commerce in new ways using stablecoin rails, to enhancing autonomous transportation, to increasing productivity and artistic expression, to curing diseases. But it is also in recognition of the risks. Blindly handing over our consequential actions to an uncontrolled intelligence can have harmful, and in some cases devastating, consequences for people and business.

Then there’s the commercial opportunity. In the U.S. alone, Morgan Stanley projects that up to 20% of US e-commerce purchases ($385 billion) will be agent-driven by 2030. To get there requires us to solve the lack of trust and misalignment of incentives that holds agentic commerce back. Amazon famously blocked Muse; at face value on the grounds of privacy, security, and violating its terms of service, but more so because bypassing Amazon’s browsing experience (designed to encourage more buying) fundamentally erodes their business model. Many smaller retailers share Amazon’s stance but will need to adapt to customer expectations that our agents will do more for us. 

Then there’s regulation. From December 2027, the EU AI Act will require AI systems used in high-risk areas, such as hiring, credit, and critical infrastructure, to keep automatic records of what they do. Companies that fall short face fines of up to €15 million or 3% of global annual turnover.

The upside to getting this right is immense.

Why Us

The first: adversarial environments are our home. We’ve spent years building in crypto, where trustlessness is the inherent value proposition, and real money depends on systems operating as intended. This is now the predominant operating paradigm of the internet: billions of agents acting for strangers, some of them trying to cheat. We designed for this world before it existed. 

The second is our technical ability. Mysten Labs is a collection of the most singular minds across programming languages, distributed systems, cryptography, and security. Sui and Walrus stand alone in their ability to coordinate at the speed and scale agentic systems demand. Our team created Move, a language built to express ownership, permissions, capabilities, and policies, exactly what agent rules require. 

The third is credible neutrality. Last week in Singapore we announced a collaboration with Google Cloud to develop Verifiable Agent Arbiter (VAA), an evidence layer for enterprise agents, built on Sui and Walrus. This is validation by one of the most important players in this industry that when agents act across trust boundaries, no single party’s logs can settle a dispute. The evidence has to be tamper-resistant and live somewhere neutral.

How We Get There

As some of you know well about us, the size of our ambitions is astronomical. To match that ambition, and to ensure that the team has full time and bandwidth to dedicate to ensure agents can trust and be trusted, we are completing two internal changes. 

One, we are formalizing the creation of the AI business unit, solely dedicated to developments of agentic trust products and AI native infrastructures. Self-contained with engineering, product development, and other functions to reduce dependencies, the AI BU will run similarly to the onchain finance BU, which we implemented earlier this summer, and which has and will continue to support a variety of product launches and partner support. 

Two, I have asked Kevin Boon, President of Mysten Labs, to step into the role of Chief Executive Officer as I add leading the AI BU to my list of responsibilities as Chief Technology Officer, and he has graciously accepted. Kevin has been here since the beginning, first as our lawyer, then as our President. He is a known quantity that has worn many hats and built trust in every facet of the company, during our crises and day to day. This makes my scope more manageable, and frees me to return to what I know I do best - being strongly opinionated with a vision for where our technology needs to go, putting us at those intersection points, and leading technical teams to make that vision real. 

Thanks for reading. For us, it’s back to building. We’ll see you next quarter. 

Evan Cheng