Jev Demo Night: 9 Days from Launch to Live Demos
Most AI models answer a question with a paragraph. Jev answers with a decision.
TypeSafe released Jev on September 15th. Nine days later, on September 24th, New York’s AI community filled Betaworks Studios in the Meatpacking District for the first-ever Jev Demo Night. Founders, developers, and tech enthusiasts came to see what people had already built with it.
The answer: a lot, in very little time.
What Jev is
Jev is TypeSafe’s first System One Model. TypeSafe is a San Francisco AI lab co-founded by Diogo Almeida, who worked on ChatGPT at OpenAI. You give Jev a structured question. It returns a typed decision and an estimate of how confident it is. Developers can use that answer directly inside of their software.
A chatbot writes. Jev decides. That difference shaped the night.
The demos
Kaya Jones from Betaworks moderated and introduced each presenter. Teams showed what they had built, the audience voted, and three winners each took home $250 in credits.
The presenters were Nick Price, Apurva Chitnis, Josh Pham, Conor Branagan, Hunter Brooks, Sam Chien, Adam Roman and Keduse Worku.
Although it was a competition, the event didn’t feel like one. The presenters all sat together at one table. The mood was supportive and collaborative, closer to a group of friends showing each other projects than rivals chasing a prize.
That says something about New York’s AI scene right now. When a new tool lands, builders here treat it as a shared puzzle first and a race second.
One standout came from Adam Roman, founder of Cosyra, who built a slop filter. He opened Instagram, then LinkedIn, and scrolled through his feed as he normally would. As he scrolled, Jev read each post. When it judged a post to be AI slop, it flagged it in red and then removed it from the feed. It all happened live, at scrolling speed. It was a simple idea, and it made the point. Jev is fast enough to sit inside something you use every day and make a call on every item without slowing you down.
The other demos covered very different ground, but they shared a theme. Each one used Jev to make fast decisions inside software, whether that meant automating a step people usually handle by hand or sorting the useful from the noise. One model was put to work on several different problems within days of its release.
Who made it happen
Betaworks hosted the night. Betaworks is a New York venture fund and startup platform that invests in early-stage companies working in AI, media and new interfaces. Its portfolio includes Hugging Face, Granola, and Sandbar. OpenRouter and LangChain sponsored the credits.
Our take
Most AI talk is about what models might do someday. This night was about what people did in a week.
Jev’s sweet spot is the small decisions software makes all day. Is this spam? Where should this request go? Does this pass the check? Those calls happen thousands of times a day and need to be fast and cheap. A chatbot is overkill for them. A model that gives one answer and a confidence score is not.
Adam Roman’s slop filter showed it best. Nobody needs a paragraph explaining why a post is AI slop. You need it gone before you see it.
The other story was the room itself. Eight builders, one new model, one table. New York’s AI community is small enough to sit together and serious enough to ship in days. That is a good combination.