&Truth, a Palo Alto and Singapore startup building what it calls a "universal verification machine" for code, said Thursday it raised a $70 million seed round as it emerged from stealth. The company announced the financing in a Business Wire release dated October 29, 2026, but declined to name its investors or disclose a valuation.
Wired reported the same day that Nvidia's venture arm is among the backers. The company's own materials name no investors, and Wired is the only outlet in the available reporting to identify one.
"Bridges, buildings, they don't collapse, because people did math before constructing them," Ilya Sergey, the company's co-founder and chief science officer, told Wired. It is a tidy analogy for a company asking the software industry to adopt a similar discipline, and one that arrives with no shipping product to test it against.
What the company is building
&Truth is a public benefit corporation, legally andTruth, PBC. It says its technology reads code in any language, proves that the code meets precise requirements, and produces proofs that anyone can check independently.
The startup also describes a "specification factory," meant to translate human and business requirements into formal specifications at scale. Formal specifications are mathematically exact statements of what a program must do, and writing them has traditionally been the slow, specialist part of the work. The company's approach rests on the Lean proof system and Mathlib, a large library of formalized mathematics. It also uses and extends CSLib, a Lean library for computer science and software.
No product has been released publicly. The company says it is building core verification technology and plans to release foundational libraries as open source, working with the Lean Focused Research Organization (Lean FRO).
Speed claims, and a pointed pitch
The company says AI accelerates formal proof work by roughly 100 times, and that its own automation adds another 100 times on top, shrinking proof efforts that once took months or years to hours or minutes. Those figures come from &Truth. The available reporting cites no independent benchmark.
The pitch leans on a sense of urgency about AI-generated code. "AI is now breaking code at superhuman speed… even 99% safe is certainly broken," said Chris Hsu, co-founder, co-CEO and executive chairman, in the release.
Sergey, in the same release, framed the company's goal as a matter of accountability: "Trust in software should rest on independently checkable proof, not any company's word, including ours."
Founders and team
Clark Barrett, co-founder and co-CEO, is a research professor of computer science at Stanford. He directs the Stanford Center for Automated Reasoning, co-directs the Stanford Center for AI Safety and leads the cvc5 project, a well-known solver used in automated reasoning. He also sits on the CSLib steering committee.
Sergey is an associate professor at the National University of Singapore, where he leads the VERSE Lab. He co-created Veil and Velvet, two Lean-based verifiers.
Hsu is president of the Infinitude Foundation and sits on the Lean FRO Strategic Advisory Board. He was previously CEO of Kilometre Capital and a managing director at Citadel Investment Group.
The rest of the leadership draws from the Lean community and from business. Sebastian Ullrich, a co-architect of Lean 4 and co-founder of Lean FRO, is chief Lean officer. Gregory Malecha, architect of BRiCk, is technical lead. Drew Harman, formerly of Ariba, is chief business officer, and Linus Lee, a former managing partner at Kilometre Capital, is chief of strategy and operations.
Potential users and hiring
Wired reported that &Truth is evaluating its technology with potential users including Nvidia and Tensordyne, an AI inference company. The release lists no customers or pilots, and the company has not given a general availability date.
It declined to disclose revenue or headcount. It said it will grow its Singapore operation across research, engineering, product and go-to-market, and its careers page is open.
Low Jin Yong, senior vice president and co-head of Digital Industry Singapore, said in the release: "As AI adoption scales, trust and safety will become increasingly important to how companies can deploy AI with confidence."
A crowded formal-verification field
&Truth enters a market that has drawn considerable investor attention in 2026. Axiom announced a $200 million Series A at a $1.6 billion post-money valuation on March 12, led by Menlo Ventures, according to Menlo. Pramaana Labs raised a $27 million seed led by Khosla Ventures on June 17, TechCrunch reported.
Theorem raised a $6 million seed, also led by Khosla, on January 27, VentureBeat reported. Qodo, which verifies AI-generated code, raised $70 million, SiliconANGLE reported on March 30.
What's next
For now, &Truth's case rests on credentials and promises rather than product. Its first public milestone will be the open-source libraries it plans to release with Lean FRO, which will give outside engineers something concrete to inspect, and perhaps to challenge.
