TypeSafe AI has officially emerged from stealth with $40 million in seed funding led by the venture capital firm DCVC. The San Francisco-based frontier lab is introducing a machine-native, composable approach to artificial intelligence that is designed for direct integration into software systems. The company was founded by Diogo Almeida, Erik Gafni, and Sasha Sheng, with Almeida previously serving as an OpenAI researcher and co-inventor of RLHF and ChatGPT.
A Founding Team With Deep Research Credentials
Diogo Almeida brings significant experience from OpenAI, where he contributed to reinforcement learning from human feedback and the development of the widely used ChatGPT system. Alongside co-founders Erik Gafni and Sasha Sheng, he is steering TypeSafe toward a vision that treats intelligence as a reliable software primitive rather than a standalone assistant. The team believes most machine intelligence will eventually operate quietly inside software, running in the background to support everyday applications and workflows.
Introducing Jev for Machine-Native Workloads
Jev is the first model built around TypeSafe's machine-native philosophy, and its name is a nod to the economic principle known as Jevons Paradox. The model delivers frontier-level intelligence at under 100 milliseconds of latency and is described as up to 100 times faster and less expensive than other frontier models. Early access is currently available to select developers, and the system can process hundreds of outputs in parallel from a single prompt while supporting high-volume software environments.
Developer Controls and Confidence Scores
One distinguishing feature of Jev is its provision of calibrated confidence scores that help developers determine when software can act autonomously and when it should defer to human review. This capability addresses a persistent challenge in production systems, where unpredictable model behavior can undermine reliability and increase operational risk. By distinguishing between high-confidence and uncertain outputs, Jev enables more controlled, testable, and consistent automation within software applications.
Rethinking Frontier Models for Production Software
TypeSafe argues that today's frontier models are optimized for reasoning and human interaction, but those strengths do not always translate well to production software environments. Models can hallucinate, change methodologies between requests, and introduce unnecessary variability into systems that depend on predictable behavior and consistent decision-making. In response, TypeSafe is building intelligence that functions as a composable primitive for applications requiring semantic judgment, classification, and intelligent decisions at scale.
Investor Confidence and Strategic Vision
James Hardiman, General Partner at DCVC, said TypeSafe is approaching one of the biggest remaining challenges in artificial intelligence by making smarter models more usable for developers. He noted that founders Diogo Almeida, Erik Gafni, and Sasha Sheng bring the technical depth and conviction needed to rethink how models are built for software. The firm believes composable intelligence can unlock an entirely new generation of applications, and DCVC led the company's approximately $40 million seed round.
Toward a Standard Building Block for Software
TypeSafe is working to augment deterministic software with AI wherever semantic judgment is needed, allowing systems to operate with greater autonomy and intelligence. The company believes that making intelligent decisions fast, affordable, and reliable enough to test, constrain, and layer upon will allow developers to create increasingly sophisticated applications over time. Its foundation is intended to enable complex software that no single company could realistically build from the top down.
TypeSafe AI is positioning itself at the intersection of frontier model research and practical software infrastructure for developers. By treating intelligence as a standard building block, the company hopes to help engineering teams integrate reliable decision-making capabilities into production systems without sacrificing speed or control. With early access waitlisted at typesafe.ai, TypeSafe is now focused on proving that machine-native AI can support a new wave of intelligent, composable applications.