TypeSafeAI positions Jev as an alternative for automated decisions

A different model for fixed decisions

TypeSafeAI unveiled Jev on Sept. 15 in Silicon Valley, according to The Wall Street Journal. Unlike text-generating chatbots and large language models, Jev uses machine learning to classify inputs into predetermined outputs, including yes-or-no answers, scores or answers from a list. The startup describes the approach as Reinforcement Learning for Calibrated Decisions.

TypeSafeAI CEO Diogo Almeida helped develop ChatGPT at OpenAI before leaving the company in 2024. He said Jev is fast, cheap and produces consistent results, positioning it as a way to address limitations in LLMs that are difficult to predict.

Claims about reliability and usage

TypeSafeAI says Jev is suited to automation software. Almeida said major LLMs handle tasks involving human intervention well but are "unbelievably useless" for automation. He compared software development with building a trusted foundation and adding features one by one, saying chatbots do not work that way.

Almeida said about 25 percent of Fortune 500 companies were already using Jev. He also said daily throughput had reached 1 trillion tokens a week earlier and described growth as exponential.

Funding and competing products

Venture capital firm DCVC invested $40 million in TypeSafeAI. The Information reported that the startup was in talks to raise more than $1 billion, with some investors citing a valuation above $10 billion. Almeida declined to comment on the discussions.

Similar services are also emerging. At the end of September, OpenAI unveiled a Decisions API powered by its Luna model that provides fixed answers to user-defined questions. Data analytics startup Databricks released an ai_decide function. Almeida expected copycat products to emerge but said he was confident Jev would be better.