Former OpenAI researcher builds an AI model that judges options instead of writing text
Back to Home
ai

Former OpenAI researcher builds an AI model that judges options instead of writing text

September 16, 2026101 views2 min read

Former OpenAI researcher Diogo Almeida launches TypeSafe AI, introducing Jev — an AI model that judges options instead of generating text, offering ultra-fast response times and low token costs.

In a bold departure from traditional AI chatbots, Diogo Almeida, a former researcher at OpenAI, has launched TypeSafe AI, a new company focused on building models that judge and classify options rather than generate text. The flagship product, named Jev, is designed to deliver precise, binary decisions within software applications, bypassing the need for natural language responses.

Minimalist AI for Precision

Unlike conventional large language models (LLMs) that produce human-like text, Jev operates on a strict classification framework. It offers no text generation, focusing solely on evaluating and selecting from preset options. This approach results in ultra-fast response times—starting at just 70 milliseconds—making it ideal for real-time software applications where speed and accuracy are paramount.

The model’s architecture emphasizes efficiency and predictability. According to TypeSafe AI, this design choice significantly reduces token costs, making it a cost-effective solution for developers and enterprises. However, the trade-off is clear: while Jev excels at sticking to predefined choices, it does not offer the flexibility or error correction that traditional LLMs provide.

Implications for AI Development

This innovation reflects a growing trend in AI development toward task-specific models that prioritize performance over versatility. By eliminating the complexity of text generation, Jev can be optimized for specific use cases such as automated decision-making, content filtering, or interface validation. While this model may not replace general-purpose AI assistants, it could carve out a niche in applications where precision and speed are more valuable than conversational fluency.

Almeida’s move underscores the industry’s shift toward more targeted, efficient AI solutions. As developers seek tools that reduce latency and enhance reliability, models like Jev may become increasingly important in building scalable, high-performance systems.

Conclusion

TypeSafe AI’s introduction of Jev marks a significant step in reimagining how AI can be used in software environments. While it may not be a replacement for chat-based AI systems, its unique approach to classification could open new possibilities in performance-critical applications.

Source: The Decoder

Related Articles