JevMade field notes / Video guide

Jev es el lanzamiento de IA mas importante desde ChatGPT

EDteam explains Jev's architecture as a probability-based decision engine rather than a generative text LLM. They cover API payload design (state, questions, criteria) across choice, Noul, and score question types, run playground experiments highlighting prompt sensitivity, and showcase an MVP course recommender sorting real catalog entries using confidence scores.

Original by EDteamGetting startedIntermediate25 min 11 sec Published 2026-09-18Source reviewed 2026-09-26

Before you press play

What you’ll find in the video

Jev structures decision queries into three core types: choice (single-selection categorical classification), Noul (binary probability between 0 and 1), and score (ranking across continuous or ordinal criteria).
A single POST request state can evaluate multiple parallel question objects simultaneously without mutual interference, returning confidence distributions and probability numbers instead of freeform text.
Prompt wording directly shifts returned probabilities, as small changes in question phrasing significantly alter calculated distribution scores in tests.

Worth knowing

Gemini-assisted video/transcript review. Presented speed, token pricing, and response latencies reflect informal vendor playground tests and unbenchmarked internal MVPs rather than independently verified production baselines.

All video guides →Prefer to read? Written guides →