TypeSafe AI, a frontier AI lab focused on building composable, machine-native intelligence, has emerged from stealth with $40 million in seed funding led by DCVC. The company was founded by Diogo Almeida, a former OpenAI researcher and co-inventor of RLHF and ChatGPT, alongside Erik Gafni and Sasha Sheng.
Almeida said TypeSafe was created to pursue an alternative research path centered on intelligence built primarily for software rather than human interaction, arguing that if AI is to reshape how work gets done, most intelligence should eventually operate quietly within systems rather than through conversational assistants. The company argues that today’s frontier models, while capable in conversation, often introduce unpredictability unsuitable for production software through hallucinations and inconsistent methodologies.
TypeSafe’s first model, Jev, named after Jevons Paradox, is designed to deliver frontier-level intelligence with under 100 milliseconds of latency, positioned as up to 100 times faster and cheaper than comparable frontier models. The model can process multiple outputs in parallel and provides calibrated confidence scores intended to help developers determine when software can act autonomously.
James Hardiman, General Partner at DCVC, said the founders’ technical background positioned them to rethink how models are constructed for software integration, describing their composable approach as a foundation for a new generation of applications. Jev is currently available through a waitlisted early access program for select developers.
