TypeSafe AI, a San Francisco-based startup, has emerged from stealth mode after securing $40 million in seed funding. The investment round was led by DCVC. The company was founded in 2024 by former OpenAI researcher Diogo Almeida, alongside Erik Gafni and Sasha Sheng.
The startup has introduced its first model, Jev, which is designed for machine-native intelligence rather than human interaction. The model aims to provide predictable outputs for software integration, moving away from the conversational focus of current frontier models.
TypeSafe claims that Jev offers frontier-level intelligence with latency under 100 milliseconds. The company states the model can be up to 100 times faster and less expensive than existing alternatives, with the ability to generate hundreds of parallel outputs from a single prompt.
The model utilizes a training method called RLCD. TypeSafe intends for developers to use Jev as a software component for intelligent decision-making within applications, rather than as a standalone chatbot.
The company's strategy relies on the economic principle of the Jevons Paradox, suggesting that significantly lower costs for AI intelligence will lead to a massive increase in its consumption across software systems.
