One round, a one-month-old product
On October 9 TypeSafe announced an $870 million Series A at a $7.5 billion valuation, led by Andreessen Horowitz with Sequoia Capital and existing investor DCVC participating; a16z partner Martin Casado joins the board. Its only product, Jev, launched on September 15 and is still in early access.
The company was founded in 2024 by former OpenAI researcher Diogo Almeida, former Meta research engineer Sasha Sheng and engineer Erik Gafni. In its funding announcement TypeSafe says a third of the Fortune 500 already use Jev and that it has "saved customers millions of dollars in production." These are company claims; no revenue or usage volume was disclosed.
What Jev sells
In agents and automated workflows, many judgments need only one option: approve this order or not, escalate this to a human or not, which tool to call at this step. Using an LLM means having it write a paragraph and then extracting the answer, which is slow and token-hungry. Jev is transformer-based but doesn't generate text: given application state and a typed question, it directly returns a probability for each option, which TypeSafe calls calibrated decisions. Almeida told TechCrunch that computers "speak a different language," so the human-language strength of LLMs isn't that useful for automation.
The followers have arrived
The space got crowded within a month: this site has covered PostHog's open 9B decision model Jeeves and AWS's open 2B Strands Decider, both modeled on Jev's interface. The difference is that Jev is available only through a hosted API, while the followers' weights can be downloaded and self-hosted.
The $7.5 billion valuation is a bet that "judgment" will be split out of LLMs into a standalone, usage-billed infrastructure layer. Developers now have two routes to compare, closed and hosted versus open and self-hosted; it's worth testing accuracy, latency and cost on one or two high-frequency decision points in your own workflow before replacing existing LLM calls.