---
title: "San Francisco falls for an AI that will not chat"
description: "TypeSafe AI’s Jev takes decisions instead of talking, and the developers — and the money — are circling"
author: "Albion Grey"
published: 2026-09-29T09:00:02Z
modified: 2026-09-29T13:36:43Z
url: https://rews.cc/a/san-francisco-falls-for-an-ai-that-will-not-chat-f58c65
language: en
tags: ["ai", "openai", "llm", "automation", "startup", "tech"]
publisher: "Rews (https://rews.cc)"
---

# San Francisco falls for an AI that will not chat

*TypeSafe AI’s Jev takes decisions instead of talking, and the developers — and the money — are circling*

By Albion Grey · September 29, 2026 · https://rews.cc/a/san-francisco-falls-for-an-ai-that-will-not-chat-f58c65

## In brief

- Jev, a decision-making AI model from startup TypeSafe AI, has gone viral, with almost 40m views of its launch video on X
- Unlike large language models, Jev returns structured answers with probability estimates rather than generating text
- Developers at a San Francisco hackathon reported weaving Jev into real products, citing big gains in speed and cost
- Venture capitalists have offered to invest in TypeSafe at a $10bn valuation, The Information reports

San Francisco’s tech scene has a ritual for new arrivals in artificial intelligence: first the hype, then the hackathon. Jev, a viral AI model from a startup called TypeSafe AI, received both this month. Its launch video has racked up almost 40m views on X since mid-September, and on Saturday developers gathered at the offices of CodeRabbit, another startup, to build with it. Business Insider’s reporter went along; the event amounted to a coming-out party.

Jev is notable chiefly for what it is not: a large language model in the style of OpenAI’s or Anthropic’s. Rather than spitting out text, it takes decisions, responding to questions rapidly with structured answers that carry probability estimates of their own accuracy. It is also, by all accounts, dirt cheap and extremely fast. Diogo Almeida, TypeSafe’s founder and a former OpenAI researcher, has argued that the industry’s obsession with chatbots does too little to automate rote work.

Allie Laabs, TypeSafe’s head of developer relations, put the heresy to the crowd directly. “What if this path that we’ve been so horse-blinders on, the LLM path, what if that’s not the only path?” she said. Her advice was to break questions into the most specific possible pieces. Jev excels at classification: mowing through customer-support queries and slotting them into buckets, for instance, or moderating a fast-moving online chat. Ms Laabs hopes its speed and low cost will tempt developers to build tools they had previously judged too ambitious.

Fueled on croissants and coffee, the developers obliged. Jalaja Kurubarahalli, cofounder of Quintess AI, which builds voice tools for maintenance workers, reckoned faster question-sorting could cut the lag in her products. Hitesh Viswasam, a robotics engineer, suggested a model like Jev could let robots respond to heat or danger at human-like speeds. Sourabh Mane, a design engineer at CodeRabbit, called it a “breath of fresh air.”

Some had gone further than tinkering. Lawrence Liang, a developer, showed off a personal AI agent into whose code infrastructure he said he had built Jev to the tune of 80%; it was his second Jev hackathon of the week. The model works out far more quickly than a large language model which tools a given query needs. “It almost blows your mind,” he said.

Avram Cheaney, an engineer, told Business Insider that adding Jev to JunkDrawer.ai, his AI search product for eBay, had made it “drastically cheaper and drastically faster” than relying on language models alone. Those few seconds matter for consumer products, whose users now expect everything to load instantly. “This is what’s been missing,” he said.

The money has noticed. Some venture capitalists have offered to invest in TypeSafe at a $10bn valuation, according to The Information — a heady sum for a firm that was anonymous a fortnight ago. Its Discord server has passed 100,000 members. “It is definitely completely nuts,” Ms Laabs said of the week, calling the reception something “very few people ever get to experience in their lives.”

She had worried that Jev’s way of working would baffle people raised on chatbots, and found that one explanation does the trick: “A software engineer has to put it in something; it isn’t anything until that.” Then, she says, the penny drops.

Whether Jev is a genuine fork in the road or merely the latest Bay Area infatuation is not a question a single hackathon can settle. The chatbots’ makers, after all, have a few billion dollars and a White House lunch habit with which to argue the other path.