FrontierSeptember 16, 2026via The Decoder
Former OpenAI researcher builds an AI model that judges options instead of writing text
Why it matters
A narrow-task model architecture challenges the generalist LLM paradigm. Practitioners evaluating cost and latency trade-offs now have a specialized alternative for high-volume classification workloads; enthusiasts see a divergence in model design philosophy emerging from the frontier labs.
Key signals
- Founder: Diogo Almeida, former OpenAI researcher
- Company: TypeSafe AI
- Model name: Jev
- Core capability: classification only, no text generation
- Response latency: starting at 70 milliseconds
- Token pricing: extremely low (specific rates not disclosed)
- Constraint: strictly adheres to preset options
- Published: September 16, 2026
- Model: Jev
- Response latency: 70 milliseconds
- Token pricing: extremely low (specific rate not disclosed)
- Trade-off: guarantees preset-option constraint; no error-mitigation for wrong classification
- Architecture: deliberately specialized, not general-purpose
The hook
Former OpenAI researcher drops a classification-only model—no text generation, 70ms latency, pennies per call.
TypeSafe AI, founded by former OpenAI researcher Diogo Almeida, is releasing a model that deliberately generates no text. Instead of chat responses, "Jev" delivers pure classifications for software, with response times starting at 70 milliseconds and extremely low token prices. The approach doesn't …