Jev, and state of AI tools

 The rise and fall of Jev - Alex Ewerlöf Notes

The article, titled "The rise and fall of Jev," uses the rapid hype cycle and quick disillusionment surrounding a new AI model called Jev as a case study to examine the current state of the AI industry. The author explores how marketing hype and tech buzzwords override critical thinking, contrasts this with four years of genuine incremental progress since ChatGPT launched, and outlines the fundamental problems that AI still has not solved.

• The Anatomy of the Jev Hype: Jev was marketed as a revolutionary "System One Model"—a fast, general-purpose classification model trained via a new method called Reinforcement Learning for Calibrated Decisions (RLCD).
  
  
• The Hype Claims: It was hyped with high-profile name-dropping, claims of being exponentially cheaper, and assertions that it "cannot hallucinate."
  
  
• The Reality Check: The hype quickly collapsed into the "Trough of Disillusionment" when users realized that while Jev outputs valid schemas (hence "no hallucination"), it can still be entirely wrong, and other open-weight models match its cost-to-performance ratio.
  
  
• Four Years of Incremental Improvements: Progress since GPT-2 and ChatGPT has been steady and practical rather than magical, introducing instruction-tuned models, function calling, coding harnesses, AI agents, larger context windows, multimodal capabilities, and Model Context Protocol (MCP).
  
  
• Why the AI Hype Persists: The industry is driven forward by fierce US-China geopolitics, a massive interdependent financial debt bubble that has to succeed, constant fear narratives around "rogue AI," and the direct disruption of white-collar and software engineering jobs.
  
  
• What AI is Still Missing: Neural networks still lack true abstract reasoning (System 2 logic), continuous real-time learning, the rapid sensory and motor precision required for embodied robotics, and the ultra-low energy efficiency of the human brain.
  
  
• Takeaways on Tech Marketing: The author advises readers to look past name-dropping, confusing jargon, ambiguity, and manufactured drama, noting that big claims require substantial, unbiased evidence.

Jev is Vercel's fastest-adopted AI model ever | TechCrunch

• Record-Breaking Adoption: Data from Vercel reveals that a decision-making AI model named Jev (created by TypeSafe AI) became its fastest-adopted AI model ever. Within two weeks of its mid-September release, it was actively used by 35% of paid teams—nearly double the reach of the second most popular model, Sonnet 5.5.
  
  
• Why Jev is Popular: Jev is a lightweight model designed specifically to choose between a few predetermined outputs rather than generating long free-form text. This makes it significantly cheaper and faster to run than standard models, making it ideal for high-volume data classification tasks (such as Vercel using it to classify harmful commands).
  
  
• Market Impact: Jev's success has inspired competing rapid-classification offerings from tech giants like OpenAI and Amazon.
  
  
• Broader Trends: Vercel's AI gateway data also highlights strong momentum for open-source models (accounting for over 70% of processed tokens), alongside rapid tenfold growth in spending on proprietary frontier labs.

Introducing System One Models & Jev - TypeSafe AI Blog