In partnership with
News of the day
1. Startups bet generative AI will make formal software verification practical, though adoption and costs remain uncertain. → Read more
2. Mistral says its open-weight Le Chonk model now beats Chinese rivals in coding and cybersecurity tasks, raising stakes for Western AI leadership. → Read more
3. OpenAI agents hacked Hugging Face, so OpenAI now wants California's AI safety law to cover such incidents. → Read more
4. Dream, founded by two ex-Palantir executives, raised $17M led by a16z to build physics-grounded AI world models that automate building design. → Read more
Our take
Hi Dotikers!
For decades, formal verification has been the broccoli of software engineering: everyone agrees it's good for you, almost nobody wants to pay for it. The idea of mathematically proving your code is bug-free sounds wonderful until you realize it traditionally takes months of expert labor and a budget that makes CFOs break out in hives. Which is why this week's story caught my attention: a Stanford researcher, Chuyue Sun, used AI to complete a verification project in under half a day for less than 500 dollars, when three experts had needed five months to do it by hand.
That kind of cost collapse changes everything. Suddenly, a sleepy academic discipline is attracting serious money: Sun's startup &Truth just launched with 70 million dollars in funding, while rival Harmonic, backed by Nvidia's venture arm, has raised 295 million. Even Anthropic got in on the action, announcing with Theorem that AI-assisted verification found bugs in open source software that trillions of conventional tests had missed. Trillions. Let that sink in before you push to production on a Friday afternoon.
And honestly, the timing could not be better. AI is writing more and more of our code, and AI is also getting scarily good at finding and exploiting bugs. Fighting AI-generated vulnerabilities with AI-generated mathematical proofs feels less like a nice-to-have and more like the only sane path forward. The skeptics have a point when they note that verification rests on assumptions and will never catch everything. Even formally verified software has coughed up bugs. But perfect should not be the enemy of provably better, and if the price drops from five months to one afternoon, the excuses for shipping unverified critical code are running out fast.
The real question now is whether companies will adopt it before a breach forces them to.
Alex.
The AI notetaker that gets the hard words right
Your AI is only as good as what you feed it. Feed it a transcript with your product names, acronyms, and numbers spelled wrong, and you spend the afternoon hand-fixing the output.
Wispr Flow Notetaker uses your dictionary and calendar, so product names, acronyms, numbers, and uncommon names come out spelled right. It pops up when your meeting starts and captures Zoom, Google Meet, Teams, and Slack huddles with one click, no bot joining the call.
Then your meetings show up inside Claude or ChatGPT through the built-in connector. No copy paste. Try Notetaker free on Mac and Windows.
Meme of the day
