Jev, an AI Model That Can’t Chat, Takes On Bigger Rivals
Read the Portuguese translation →
Author: Bloomberg Television · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Jev, an AI Model That Can’t Chat, Takes On Bigger Rivals · Channel: Bloomberg Television · 00:33 · Watch on YouTube
Topics: Data, databases and embeddings, Model inference and optimization, Markets and economy, AI models
Summary
Section titled “Summary”San Francisco-based Type Safe AI has drawn attention with Jev, a model developed by a startup founded by a former OpenAI researcher. Unlike general-purpose language models, Jev does not generate text or code; it evaluates unstructured data, selects a predefined response, and assigns a confidence score. The approach is intended to reduce costs by avoiding token-intensive text generation, and investors are reportedly eyeing a valuation above $10 billion.
Key points
Section titled “Key points”- Type Safe AI was founded by a former OpenAI researcher.
- The San Francisco-based startup went viral with its Jev model.
- Investors are eyeing a valuation of more than $10 billion.
- Jev is designed for developers managing unstructured data.
- Unlike larger models from OpenAI and Anthropic, Jev does not write code or generate text.
- Jev evaluates information, chooses a predefined response, and assigns a confidence score.
- The model’s approach is presented as less expensive than broader-purpose models from OpenAI, Anthropic, and Google.
- The video compares Jev’s low-cost approach to Deep Seek’s disruption in 2025.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | Type Safe AI and Jev | The video introduces Type Safe AI, its viral Jev model, and investor interest in a valuation above $10 billion. |
| 00:00 | A different kind of model | Jev is designed for developers working with unstructured data and returns predefined responses rather than generating text or code. |
| 00:33 | Lower-cost approach | By skipping text generation, Jev aims to use fewer tokens and cost less than general-purpose models. |
Quotes
Section titled “Quotes”This is a model offering a faster, cheaper alternative to traditional large language models. — 00:00
It’s designed for developers managing unstructured data. — 00:00
picks a predefined response, and then assigns [music] a confidence score for that answer. — 00:33
By skipping text generation, which can use a lot of tokens, Jev drastically cuts costs — 00:33
Concepts and entities
Section titled “Concepts and entities”Jev, TypeSafe AI, OpenAI, Anthropic, Google, DeepSeek, dados não estruturados, Pontuação de confiança
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 An AI startup founded by former OpenAI researcher is taking Silicon Valley by storm. San Francisco-based Type Safe AI went viral with Jev. This is a model offering a faster, cheaper alternative to traditional large language models. And investors are already eyeing a valuation of more than $10 billion. Unlike bigger models from the likes of OpenAI and Anthropic, Jev doesn’t write code or generate text. It’s designed for developers managing unstructured data. So, it simply evaluates information,
00:33 picks a predefined response, and then assigns [music] a confidence score for that answer. That approach makes Jev less expensive than general-purpose models from the likes of OpenAI, Anthropic, and Google, which are designed for broader tasks and are [music] considered token hungry. By skipping text generation, which can use a lot of tokens, Jev drastically cuts costs, drawing comparisons to Deep Seek’s low-cost disruption back in 2025.