&Truth Emerges With $70M Seed for Software Verification

Building a mathematical vocabulary for software is a substantial investment before a customer ever asks for a proof. &Truth, the Palo Alto and Singapore company developing automated code verification, is funding that work with $70M in seed capital disclosed in its October 29, 2026 emergence-from-stealth announcement.

Founded in early 2026, &Truth brings together co-founders Chris Hsu and Clark Barrett, both co-CEOs, and chief science officer Ilya Sergey. The public benefit corporation is developing technology to check whether legacy and AI-generated software satisfies precise requirements, with mathematical evidence that others can independently inspect. The announcement leaves the investors unnamed and does not disclose a valuation.

The commercial ambition reaches beyond generating proofs faster. &Truth wants to build reusable foundations for expressing how software behaves across programming languages, together with technology for translating business intent into formal specifications. For enterprise buyers, that puts the definition of acceptable behavior inside the product conversation: a powerful proof remains attached to the requirement somebody chose.

Why software needs a formal vocabulary

&Truth's technical premise starts with an awkward translation problem. Mathematics already works with carefully defined objects, while a software business has evolving code, different languages and people describing requirements in ordinary speech. Before a verification system can establish a property, that property and the relevant software behavior need a precise representation.

Lean provides an open-source programming language and proof assistant, giving researchers a system for expressing and checking formal arguments. CSLib, the Lean Computer Science Library, aims to formalize computer-science theories and tools and create common foundations that different projects can reuse. Those are underlying open-source resources, rather than independent evidence that &Truth has already solved verification for every language.

The economic attraction is reuse. Building a model of a programming language or a reusable proof technique can serve more than the first piece of code that motivated it. That gives the company's capital a different assignment from simply adding more engineers to a review queue: part of the work is constructing foundations that subsequent verification jobs can draw on.

&Truth says it intends to release foundational libraries as open source. That choice could let other researchers inspect, improve and build on the same vocabulary, while leaving the company to develop a commercial product around applying it. The company's announcement establishes that intention; it does not supply a complete language-coverage matrix, release timetable or independent benchmark of the finished platform.

The specification becomes part of the product

The company's proposed specification factory addresses the next translation: moving from what a business wants into a statement that a verification engine can use. That work is consequential because a requirement can be expressed precisely and still fail to capture what the buyer actually needs. The formal statement itself therefore deserves review alongside the implementation.

Consider an illustrative spending-control rule for an AI agent. A business might care about an aggregate budget, which users can authorize a purchase and whether a recurring charge counts against the same limit. A verification exercise needs an explicit property to check, along with assumptions about the surrounding system. Those choices carry operating meaning even before any proof is generated.

This is where product managers, security leaders and domain experts enter the commercial story. They understand the obligation the software is meant to carry. As requirements change, a buyer would also need to understand which specification applies to which version of the code, and what happens to the evidence when either changes. These are adoption considerations, rather than disclosed &Truth customer practices.

In his Ground Truth essay, Chris Hsu places human judgment at the center of deciding what systems should guarantee. The practical consequence is a continuing responsibility for the buyer: inspect the property, understand its assumptions and keep its meaning connected to the business. Formal verification makes that responsibility more explicit as software production accelerates.

Researchers and operators share the assignment

&Truth's current leadership team puts formal-methods expertise alongside commercialization roles. Clark Barrett co-leads technical work as co-founder and co-CEO. His Stanford profile identifies him as the Corbel Professor (Research) of Computer Science and describes his work in automated reasoning, formal verification and the safety of computer systems.

Ilya Sergey, co-founder and chief science officer, is an associate professor at the National University of Singapore and leads its Verified Systems Engineering lab. Chris Hsu combines the co-CEO and executive-chairman roles. The company identifies his earlier work at Kilometre Capital and Citadel, placing an investment and operating background alongside the academic founders' research experience.

The wider team includes technical lead Gregory Malecha, chief Lean officer Sebastian Ullrich, chief of strategy and operations Linus Lee, and chief business officer Drew Harman. Their responsibilities make the translation effort tangible: proof technology, language infrastructure, operations and commercialization all have work to do before a mathematical capability becomes an enterprise purchase. No CTO is named in the reviewed company materials.

&Truth has its U.S. headquarters in Palo Alto and international headquarters in Singapore. The announcement describes coordinated engineering and product work in both locations and planned Singapore expansion across research, engineering, product and go-to-market. Its careers page also identifies forward-deployed engineering and developer experience, connecting the research program to the work of bringing verification into customer environments.

What the seed capital puts in motion

The disclosed financing is $70M of seed capital. The October 29 announcement establishes the public disclosure, rather than an independently verified date on which every dollar closed. It does not identify the individual backers, a lead investor, valuation, detailed allocation of proceeds or an earlier funding history. The official text uses the dollar symbol without expressly stating a denomination.

The development ambition is broad, covering legacy and AI-written software across languages. &Truth also makes substantial claims about accelerating proof generation, but the reviewed announcement provides no independent benchmark with which to assess those gains. For a prospective engineering buyer, examples of supported code, the properties being proved and the assumptions behind those proofs would be more useful diligence material than a headline multiplier.

The company's public benefit structure adds a stated purpose around safe AI development. Its proposed use of independently checkable proofs gives the buyer a concrete way to question a vendor's assurance: examine the evidence and the property to which it applies. That mechanism still leaves requirements, surrounding systems and deployment choices inside the engineering conversation.

As &Truth builds its libraries and expands its team, the enterprise sales discussion will have to travel between those layers. A researcher can explain the formal model; a security leader can identify the property that matters; a product owner can explain why the business needs it. The company's Singapore and Palo Alto teams are assembling the machinery that would let those people keep working from the same statement as the code changes.