A new Palo Alto company wants to make "trust me, the tests passed" obsolete. On October 29, &Truth came out of stealth with $70 million in seed capital to build what it calls a "universal verification machine," a system meant to read code in any programming language and mathematically prove it is safe, secure and correct, with proofs that anyone can check independently. For founders and engineers who now ship a large share of their code with AI assistants, this is one of the more interesting bets of the fall, because it goes after the review bottleneck that every fast-moving team has started to feel.
Who is behind &Truth
The founding team reads like a formal methods all-star roster. According to the company's announcement, co-founder and co-CEO Clark Barrett is the Corbel Professor of Computer Science at Stanford, the founding co-director of the Stanford Center for AI Safety, director of the Stanford Center for Automated Reasoning and the project lead for the cvc5 automated reasoning engine. Co-founder and Chief Science Officer Ilya Sergey founded the VERSE Lab at the National University of Singapore. Chris Hsu, the third co-founder, serves as co-CEO and executive chairman and sits on the Lean FRO Strategic Advisory Board.
The wider team includes Sebastian Ullrich, co-architect of Lean 4 and co-founder of Lean FRO, and Gregory Malecha, who previously led formal methods at BlueRock Security. The company was founded in early 2026, is set up as a public benefit corporation, and has its U.S. headquarters in Palo Alto and its international headquarters in Singapore. The release describes its backers only as "a group of formative AI safety and venture capital leaders" and doesn't name a lead investor, so treat any list of investors you see elsewhere with caution until the company confirms it.
What the product actually does
Formal verification means writing a precise specification of what software should do and then producing a machine-checkable proof that the code meets it. The idea is decades old, and it has been used for things like operating system kernels, compilers and cryptography libraries, but it has always been slow and expensive because proofs were written by a small number of specialists over months or years.
&Truth's pitch is that AI changes the economics. The company says AI speeds up proof work "by 100x or more" and that its own proof automation adds another 100x, which would shrink proof generation "from years or months to hours or minutes." Those are company claims, and nobody outside &Truth has published numbers to back them yet. The company is also building what it calls a "specification factory," which turns plain business requirements into formal specifications, because writing the spec is often the hardest part of the whole process.
The stack leans heavily on Lean, the open-source proof assistant, and on Mathlib, the community library that the company says contains more than 2 million lines of mathematical definitions and theorems. Barrett put the logic plainly in the announcement: "AI transformed mathematics because Lean and Mathlib gave it a precise language for reasoning." &Truth says it plans to release its foundational libraries as open source, and Sergey framed that as a principle: "Trust in software should rest on independently checkable proof, not any company's word, including ours."
Why this matters for Bay Area founders and engineers
The problem &Truth names in its release will sound familiar to anyone running an engineering team right now: "AI-generated code is being deployed faster than people can review it, while AI-powered attackers increasingly exploit vulnerabilities in the legacy software." If your team already uses coding agents in the way we described in our guide to AI workflows for work, you've probably noticed that writing code got cheap while reviewing it stayed expensive. The company's long-term goal is to move the industry from "vibe-coding" to what it calls "veri-coding," where the AI that writes the code also hands you a proof.
For founders, there are 3 practical takeaways. First, verification is becoming a product category of its own, which means security, fintech, healthtech and infrastructure startups may soon get asked by customers how they check AI-written code, the same way they get asked about SOC 2 today. Second, the skills gap is real, since people who know Lean, Coq or automated reasoning have always been scarce, and engineers who pick up those tools now will be in demand. Third, as more of the software stack gets written by agentic AI systems, the companies that can prove their code is correct will have a sales advantage that's hard to copy.
There is also an open-source angle worth watching. If &Truth ships its libraries openly, it fits the same pattern that has pushed open-weight models forward, where shared tools let small teams build on work that used to sit inside a few large labs. That topic came up at our Open Weights AI Summit with Aaron Levie in Redwood City back in September, and it's a big reason why so many builders on the Peninsula care about who controls the tooling.
A Palo Alto story with a Stanford core
&Truth is another sign that the deep tech scene around Stanford and Palo Alto is active again, with research groups spinning out companies that tackle hard technical problems. If you're a founder or engineer down here who wants to meet the people working on AI safety, automated reasoning and infrastructure, South Bay Renaissance, the community Tribe organizes to rebuild the South Bay and Peninsula startup scene in person, runs AI deep dives, founders' dinners and demo nights from Palo Alto to San Jose. You can also see more ideas in our roundup of where to meet founders in the South Bay and Palo Alto.
What to watch next
The questions for the next few months are pretty concrete. Will &Truth name its investors and early design partners, and which industries will sign up first? Will the open-source libraries arrive, and will they work with the Lean ecosystem that researchers already use? And will the 100x claims hold up when outside teams try the tools on real production code? Those answers will decide whether verification becomes a normal step in every CI pipeline or stays a niche for chip designers and cryptographers.
If you want to talk through what this means for your team's stack, the AI Trends Tribe is where our members trade notes on coding agents, model releases and the tools that actually stick. And if you'd rather have these conversations over dinner, follow the Tribe Events calendar on Luma for upcoming AI deep dives, founder dinners and demo nights around the Bay Area.
