The Pulse
&Truth Raises $70 Million to Automate Code Proofs
&Truth emerged from stealth on October 29 with $70 million in seed funding to automate formal verification of software. Its system will use Lean to produce independently checkable proofs that code meets precise specifications, the Palo Alto and Singapore company says.

On October 29, &Truth emerged from stealth with $70 million in seed funding and a plan to automate mathematical proofs about software. &Truth’s system will read code in any programming language and prove that it meets stated requirements, with evidence others can check, the public-benefit company says. It is building teams in Palo Alto and Singapore, but the announcement describes a broad technical ambition rather than a generally available product.
Source businesswire.com
Lean is the base for a new software vocabulary
Formal verification uses mathematical reasoning to show that a program satisfies a specification, rather than relying only on tests of selected cases. &Truth plans to build on Lean, an open-source proof system, and CSLib, an open-source project developed by members of its team. Its stated work includes mathematical models of programming languages and systems, proof methods and precise ways to describe the behavior software should have.
That last step is central: a proof can establish that code meets a specification, but it cannot establish that the specification captures what a person actually meant. &Truth is developing what it calls “specification factory” technology to turn business requirements and human intent into formal specifications. Clark Barrett, the company’s co-founder and co-CEO, said, “Once we can state precisely what code must do, AI can help prove that it does it, at a speed and scale previously out of reach.”
The speed figures are company claims
AI can accelerate proof work by 100 times or more, and &Truth’s own automation adds another 100-fold improvement, the company says. The combined effect could reduce work that once took months or years to hours or minutes, according to the company. Those are claims in the launch announcement; it does not identify customers or provide a benchmark demonstrating the claimed reductions.
WIRED’s October 29 report placed &Truth among a wider group of startups trying to make formal verification easier to use, but described the technology as still at the prototype stage. Companies also remain uncertain about pricing and how customers will adopt these tools, the report said. Will Wilson, who runs the software-testing startup Antithesis, told WIRED that formal verification can be “a trap” for its practitioners; his company found additional errors after spending millions of dollars formally verifying some cybersecurity defenses.
A proof depends on its assumptions
A mathematical proof can show that code meets specified requirements across all possible inputs, under clearly stated assumptions, &Truth says. That is a narrower claim than proving software is entirely free of defects: the result depends on the requirements, the model of the system and the assumptions built into the proof. WIRED also reported concerns from formal-verification specialists that AI-generated specifications could create unjustified confidence if they encode the wrong intent.
&Truth intends to release its foundational libraries as open source, so outside researchers and developers can inspect and use them. Its launch statement does not name the investors behind the seed round, describing them only as leaders in AI safety and venture capital. Singapore will serve as its international headquarters, and &Truth plans to expand the office across research, engineering, product and commercial work.
Sources
- Original source · Press release Business Wire businesswire.com
Other pages read
Related stories
- Axiom Formalizes the 15,000-Page Classification of Finite Simple Groups
- Nous Research raises $90 million at a $1.5 billion valuation
- MAGS Verifies Agent-Written Code, Then Exposes Its Limits
- Claude Turns Wiles’s Fermat Proof Into 13 Million Lines of Lean
- TypeSafe Has Raised $870 Million for Jev, Its Non-Chat AI Model