Jev AI: Thinking Fast with real system one problems

Jev is a new kind of AI model from TypeSafe AI: instead of generating text, it returns structured, calibrated decisions.

In this video, I look at what Jev is, how its “System One” approach differs from a general-purpose LLM, and where a model like this could fit inside software systems.

The important caveat: constrained output is not the same thing as correct output. Jev may be unable to produce malformed answers, but it can still choose the wrong valid answer with high confidence.

In this video

• 0:00 What is Jev AI?
• 0:51 Thinking Fast, Thinking Slow
• 4:05 Jev use case 1: Prompt injection detection
• 5:37 Jev use case 2: Bug/Issue triage
• 8:03 Jev use case 3: AI in gaming
• 10:05 My GitLab todo prioritizer
• 14:31 Outro