Cajal

Cajal is taking the brutally strict logic of formal verification — the kind that's been securing芯片 and aerospace systems for decades — and pointing it at AI, which is honestly the most chaotic thing to happen to computing since the semicolon. They're starting in quantum computing and finance, which is a smart move because both fields are absolutely drowning in edge cases where "probably right" just isn't good enough, and their approach of grounding AI outputs in mathematically proven foundations is genuinely different from everyone else throwing "reasoning" in the name of their model. The real test will be whether they can make formal methods feel accessible enough that quantum hardware teams and trading desks actually adopt them — rigor is only useful if people use it.