&Truth reportedly raises $70M to automate formal software verification

2 weeks ago 31

The startup is betting that automation can make mathematical bug-proofing cheap enough for everyday software teams

Software developers have long dreamed of proving, with actual math, that their code has no bugs. &Truth has reportedly raised $70 million to turn that dream into something companies can buy.

The startup wants to automate formal verification, according to Wired. The technique is powerful but rarely used.

Why proving code is so hard

Formal verification works differently from ordinary testing. Testing checks how a program behaves in situations engineers think to try. Formal verification aims to show, mathematically, that the program behaves correctly in every situation.

The work has historically been slow and painful. Engineers must first write precise statements describing what their software is supposed to do. They then translate the code into formulas that a machine can check against those statements.

Each step demands specialized skill and lots of time. Most companies decide the bill isn’t worth paying. They run sophisticated test suites to catch as many flaws as possible and accept the leftover risk.

&Truth is reportedly targeting that gap with its $70 million raise. The pitch is simple: if a machine can handle the tedious translation work, formal verification stops being a luxury.

A crowded field is forming

&Truth would not be alone. Several startups have raised money recently to apply AI to formal proofs. Large language models are the engine behind most of these bets.

Theorem raised a $6 million seed round in January 2026, led by Khosla Ventures. The company is building AI tools that generate formal proofs for software.

Pramaana Labs followed in June 2026 with $27 million, also led by Khosla Ventures. It is focused on similar work for regulated industries.

VerifAIX raised $5 million in September 2026 to improve formal verification methods. Harmonic was expanding its formal verification checks in late October 2026.

The most attention-grabbing data point came from a collaboration between Anthropic and Theorem in October 2026. The work showed large language models producing formal proofs for security properties across more than 100 real-world libraries.

Those proofs reportedly surfaced bugs that had slipped past trillions of tests.

A reported $70 million round would make &Truth’s raise larger than the rounds for Theorem, Pramaana Labs and VerifAIX combined.

From research labs to production

Formal verification is not new. The concept has lived mostly in universities and in a small set of high-stakes engineering projects where failure was unacceptable.

What has changed is the cost of the labor. The hardest part of formal verification has always been the human work of writing specifications and proofs. Large language models are now showing they can take on chunks of that work.

What this means for software teams and investors

For engineering leaders, the main question is economic. Formal verification has always offered stronger guarantees than testing. It lost the argument on price and speed, not on quality.

Teams in finance, healthcare and other regulated fields could gain the most. Those industries already spend heavily on compliance, and a provable correctness guarantee is easy to justify there. Pramaana Labs has explicitly targeted those industries.

For investors, the funding pattern points to a category taking shape. Khosla Ventures has now led rounds in two companies in the space. Rounds ranging from $5 million to a reported $70 million suggest that the market is still sorting out who leads and how much the leaders are worth.

There are real risks. A formal proof is only as good as the specification behind it. If the description of what software should do is wrong or incomplete, the math can confirm a flawed design with total confidence.

The Anthropic and Theorem results offer one kind of evidence. Catching bugs missed by trillions of tests across more than 100 libraries is the sort of benchmark that gets engineering managers to pay attention.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.

Read Entire Article