TypeSafe AI: who makes Jev, and what else it builds

TypeSafe AI is the company that makes Jev, the model that answers questions about text with a typed value and a probability instead of writing text back. It describes itself as an AI lab "building intelligence beyond chat": models designed to work inside software and power automation. Its office is in San Francisco, and its jobs page lists only in-person roles.

The company came out of two years in stealth on 15 September 2026, the day Jev launched. A $40 million seed round led by the venture firm DCVC was announced the same day. Twenty-four days later, on 9 October, TypeSafe announced an $870 million Series A.

• What is Jev?
• TypeSafe raises $870M: what it means for Jev

Diogo Almeida is the founder and chief executive. He wrote TypeSafe's launch post, where he says that at OpenAI he "helped build the methods that made language models useful at following instructions". That is his own account of his work. Press coverage describes him as a former OpenAI researcher.

Two co-founders are named in reports of the seed round: Erik Gafni and Sasha Sheng. Dealroom and Value Add VC both give those names and both point back to the company's press release. We could not open that release, so treat the names as reported rather than confirmed by us. We found no sourced job titles or past employers for them, so we give none.

One model family and the tools around it. TypeSafe calls its models System One models: models built to be called by software, not chatted with. Jev is the first one it has released publicly. TypeSafe says it trains them with a method it calls Reinforcement Learning for Calibrated Decisions, so that the confidence each answer carries means something.

You reach Jev through TypeSafe's API, with documentation and a web console on the company's own domain. Around it TypeSafe publishes Python and JavaScript SDKs, agent skills and an n8n node as open source, while the model itself stays closed.

• Is Jev open source?
• Jev API access: five ways to call it
• Jev on GitHub: official repos and top projects

The funding is covered in full in our Series A article. In short: $40 million seed led by DCVC, then $870 million led by Andreessen Horowitz at a reported $7.5 billion valuation, with Sequoia Capital also taking part.

On customers, TypeSafe names one in its own materials. Its case study says the hiring marketplace Jack & Jill replaced a Google model for the first scoring stage of candidate matching and cut the cost of scoring 1,000 candidates by 88%, from $0.755 to $0.092, with screening about twice as fast. Those figures come from TypeSafe and Jack & Jill, and the company says the quality difference was not statistically significant. TypeSafe also says about a third of the Fortune 500 use Jev; Andreessen Horowitz's own post says 25%. Neither names the companies.

TypeSafe's documentation has a legal page. What it says, and what it leaves out:

• No training on your data. The page states a "commitment not to train models on user data".
• Zero data retention. TypeSafe offers it to enterprise customers, who can ask through its support form.
• Documents. The page links a Data Processing Agreement, a Master Customer Agreement and a Privacy Policy. The company's site also has terms of use, an acceptable use policy and an accessibility page.
• What we did not find. The legal page we read does not mention security certifications such as SOC 2 or HIPAA support. Ask TypeSafe if you need them.

• Jev limitations

TypeSafe's contact page offers three routes, Enterprise, Partnerships and Individual Builders, plus a support center. It lists no email address.

On 29 October 2026 its jobs page listed 15 open positions, all in San Francisco and in person, 14 full-time and one six-week contract. They include model, platform and console engineers, a developer advocate, a product manager, a recruiter and a head of legal. That tells you the company is still building out its team.

No. Jev Hub is an independent guide to Jev. It is not run by TypeSafe, it is not paid by TypeSafe and it does not resell Jev access. Other sites use similar names; the company's own addresses are typesafe.ai, docs.typesafe.ai and console.typesafe.ai, with the GitHub organisation typesafe-ai and the X account @typesafeai.

TypeSafe is about six weeks old in public and very well funded, with a short public record: seven blog posts, one named customer case study, and figures it presents itself. That is not a reason to avoid Jev, which you can test cheaply on your own data, but it is a reason to read the company's own numbers as claims and to pin a model version rather than rely on a moving alias.

• Jev pricing