The ‘Pain in the Ass’ Math That Could Save Vibe Coding From Itself — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Chuyue Sun reported in a recent paper that an AI tool completed an engineering project in less than half a day for under $500, versus five months of manual work by three highly skilled experts; her Stanford research used AI to verify part of Signal's message-encryption code.
- &Truth publicly launched on Thursday after raising $70 million, with Sun joining in August; chief science officer Ilya Sergey argues the technology for reasoning mathematically about code already existed and merely needed to scale.
- Harmonic, co-founded by Robinhood CEO Vlad Tenev, has raised $295 million and says its Aristotle agent supports verifying software written in C and hardware designs in Verilog; potential users include chipmakers Tensordyne and Nvidia, whose venture arm is an investor.
- Theorem and Anthropic said AI-assisted verification found bugs in open source software that trillions of tests had missed, and introduced a Claude add-on for other developers; the initial project's costs were not disclosed.
- Antithesis spent millions of dollars formally verifying some of its cybersecurity defenses — a process Will Wilson calls "a pain in the ass" — and still found additional errors with its own tool, plus bugs in software several customers had formally verified.
- Proofcraft says interest from potential customers now outpaces what it and its competition can provide, while CEO June Andronick and chief scientist Gerwin Klein worry AI-generated specifications could inspire "unjustified trust" in the results.
- Beneficial AI Foundation used an Anthropic grant to validate the curve25519-dalek encryption library, the same code Sun verified at Stanford; maintainer Michael Rosenberg has "no ideas" how ongoing verification will be funded and says it shouldn't rely on AI companies' goodwill.
Why it matters: Security-critical software already relies on proofs — Boeing planes, Amazon servers and more than 2.5 billion active Apple devices include formally verified code, and Apple says corecrypto issues were undetectable otherwise — but adoption is gated by review time: midsized firms typically verify one key feature or none, citing weeks-long waits and scarce verification engineers.
Ask SkimNews
