{
  "id": 12935951,
  "title": "Jev, an AI for making quick decisions, has been a viral hit in Silicon Valley. But OpenAI is hot on its heels",
  "url": "https://urgent.news/2026/10/08/jev-an-ai-for-making-quick-decisions-has-been-a-viral-hit-in-silicon",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-08T18:38:30.000Z",
  "source": {
    "name": "Fortune",
    "slug": "fortune",
    "url": "https://fortune.com/2026/10/08/jev-an-ai-for-making-quick-decisions-has-been-a-viral-hit-in-silicon-valley-but-openai-is-hot-on-its-heels/"
  },
  "original_language": "en",
  "account": "Jev, an AI model developed for rapid decisions, gained viral popularity among AI developers and Silicon Valley insiders when TypeSafe AI launched it in September. Early users quickly experimented with Jev's capabilities, which included sorting online customers into categories, categorizing documents, and identifying writing errors—all at a lower cost than many other AI models. Within a day of being available on Vercel’s AI Gateway, nearly 13% of Vercel’s paid teams had tested Jev, which had achieved this level of adoption faster than any previous model launch. OpenAI responded just three weeks later with Decisions API, a competing service that employed GPT-6 Luna, an existing model in its lineup. Both companies aimed to make AI practical for routine judgments businesses might automate. TypeSafe trained Jev specifically for this purpose, whereas OpenAI's Decisions API utilized an existing model. Jev's creator, TypeSafe CEO Diogo Almeida, had previously worked at OpenAI and observed a gap between models' ability to answer questions and their utility in automating routine work. Almeida aimed to address AI's \"massive over-promise under-deliver\" issue and prevent an \"AI winter.\" He reasoned that most requests to models would likely come from code rather than people, leading to the development of Jev. Jev's classification tasks involve determining whether incoming messages pertain to billing or technical issues, routing them accordingly based on a confidence threshold. While rudimentary AI systems have handled such classification tasks for years, LLMs offer finer-tuned decisions but at higher costs and slower speeds. Jev aims to balance these factors, enabling developers to set up classifiers through easy natural language instructions without the need for extensive programming expertise. However, developers still need to verify Jev's accuracy for specific tasks.",
  "summary": "After helping develop ChatGPT, TypeSafe CEO Diogo Almeida spent two years building Jev, which went viral with developers. His former employer is now racing to catch up.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}