TypeSafe AI
An AI startup founded by former OpenAI researchers, focused on developing decision models. Its products are designed for rapid-judgment scenarios and support probability and confidence score outputs.
Core Facts
Jev is a type of front-end model called a 'System 1 model' — fast, cheap, and structured, designed specifically for decision-making; it does not generate text but instead returns probability scores within a predefined set of options.
Jev's response time is 70–500 ms.
Timeline (last 90 days)
The paper's motivation comes from decision models such as Jev (TypeSafe AI, 2026), which return a probability for any yes/no question a developer writes down.
RuntimeWire, citing Bloomberg, noted that TypeSafe's financing round did not disclose a closing date.
Crypto Briefing added that some investors had previously discussed a financing size exceeding $1 billion, at a valuation above $10 billion.
Dealroom classifies TypeSafe's round as a late-stage round rather than a Series A, and ranks it in the 97th percentile of late-stage deep tech rounds in the United States.
According to Pulse 2.0, TypeSafe's funding from this round will be used to expand Jev and build more models aimed at machines rather than chat, while scaling up infrastructure and adding enterprise-grade features in security, governance and integration.
TypeSafe said on October 7, 2026 that recruiting company Jack & Jill cut screening costs by 88%, from $0.755 per 1,000 candidates to $0.092, and median screening time from 20.3 seconds to 10.3 seconds.
Dealroom reported that TypeSafe claims Jev has surpassed 1 million users.
Jev is not a traditional chatbot; it receives the program's current state and returns typed values conforming to a developer-defined format, an architecture TypeSafe calls a "System One model".
TypeSafe announced on September 15, 2026 a $40 million seed round led by DCVC, at a valuation of about $200 million at the time.
TypeSafe AI announced on October 9, 2026 the completion of an $870 million Series A financing at a valuation of $7.5 billion, led by Andreessen Horowitz (a16z), with Sequoia Capital, DCVC and angel investors participating.
38 more timeline events
All Facts (20)
Jev is a type of front-end model called a 'System 1 model' — fast, cheap, and structured, designed specifically for decision-making; it does not generate text but instead returns probability scores within a predefined set of options.
75%VerifiedJev's response time is 70–500 ms.
65%UnverifiedJev is positioned as TypeSafe AI's 'System One' frontier model, focusing on scenarios within software systems that require fast, structured judgment.
60%UnverifiedThe paper's motivation comes from decision models such as Jev (TypeSafe AI, 2026), which return a probability for any yes/no question a developer writes down.
50%UnverifiedDealroom reported that TypeSafe claims Jev has surpassed 1 million users.
50%UnverifiedJev is not a traditional chatbot; it receives the program's current state and returns typed values conforming to a developer-defined format, an architecture TypeSafe calls a "System One model".
50%UnverifiedAccording to Pulse 2.0, TypeSafe's funding from this round will be used to expand Jev and build more models aimed at machines rather than chat, while scaling up infrastructure and adding enterprise-grade features in security, governance and integration.
50%UnverifiedDealroom classifies TypeSafe's round as a late-stage round rather than a Series A, and ranks it in the 97th percentile of late-stage deep tech rounds in the United States.
50%UnverifiedTypeSafe said on October 7, 2026 that recruiting company Jack & Jill cut screening costs by 88%, from $0.755 per 1,000 candidates to $0.092, and median screening time from 20.3 seconds to 10.3 seconds.
50%UnverifiedRuntimeWire, citing Bloomberg, noted that TypeSafe's financing round did not disclose a closing date.
50%UnverifiedCrypto Briefing added that some investors had previously discussed a financing size exceeding $1 billion, at a valuation above $10 billion.
50%UnverifiedThe TypeSafe team revealed that just a few weeks after launch, 25% of Fortune 500 companies were already using Jev.
50%UnverifiedTypeSafe AI's Jev model accepts a state and a set of typed questions, and directly returns calibrated probability values without requiring natural language text parsing.
50%UnverifiedTypeSafe AI's team includes COO Sasha Sheng from Meta/FAIR and serial entrepreneur CTO Erik Gafni, with the overall team drawing members from OpenAI, Google Brain, and Meta/FAIR.
50%UnverifiedTypeSafe AI stated that approximately one-third of Fortune 500 companies have begun using or integrating Jev.
50%UnverifiedTypeSafe AI was valued at $200 million in its seed round in mid-September, a 37.5-fold increase compared to its Series A valuation.
50%UnverifiedThe update normalizes all publicly exposed state forms into a parts array before invoking the decision provider, with text presented as text parts and JSON objects as json parts.
50%UnverifiedArrays passed directly as state are now treated as parts of the decision state; if a JSON array needs to be passed as data, it should be wrapped in an object or explicitly wrapped as a json part.
50%UnverifiedWhen the state contains exactly one JSON part, TypeSafe AI preserves the native JSON state form rather than forcibly decomposing it.
50%UnverifiedThe model call span records the normalized state parts, while the outer span retains the publicly exposed state input.
50%