Twenty-four days ago, almost nobody had heard of TypeSafe AI. Now investors are putting a $7.5 billion price tag on it — and this time the hype isn't about another chatbot.
The startup, founded in 2024 by former OpenAI researcher Diogo Almeida, ex-Meta engineer Sasha Sheng and entrepreneur Erik Gafni, built a model called Jev. It runs on the same transformer architecture as GPT or Claude, but it doesn't write sentences. Jev outputs probabilities — what the company calls "calibrated decisions." Almeida's pitch is simple: language models spent four years getting good at human speech, but that's useless for automation, because machines don't need words, they need numbers.
Jev launched on September 15 and crossed a million users within days. TypeSafe says a third of Fortune 500 companies already run it, mostly for tasks where speed and compute cost matter more than fluent conversation. Andreessen Horowitz led the $870 million round, with partner Martin Casado taking a board seat; Sequoia Capital and DCVC also joined in.
The investor logic is straightforward. Most corporate automation isn't a dialogue — it's millions of small decisions a second: approve a transaction, route a request, flag an anomaly. Running a full language model for that is slow and expensive. A model built specifically for probabilistic output can do the same job for a fraction of the cost, and that's exactly what Jev is selling enterprise clients.
The open question is whether real revenue can keep pace with that valuation. Going from launch to $7.5 billion in 24 days is a speed only a handful of AI companies have ever hit, and some of them later fell short of what investors paid for. A million signups in a few days says nothing about how many of those users actually pay, or whether they stick around a year from now — and that's what will ultimately decide if the price tag holds up.
