Docs

What is the Jev model?

Jev is the TypeSafe AI model that scores PostLens drafts — what it is, how PostLens uses it, and what it is not.

Jev is a model from TypeSafe AI built to answer structured, closed questions and return calibrated distributions rather than free text. PostLens uses it as the scoring engine behind every writing signal.

How PostLens uses Jev

One draft becomes one Jev request carrying eleven closed questions — eight rubric factors on a 0–4 scale, one genre choice, one bait check, and one generic/formulaic check. Jev returns a level and a probability distribution for each. PostLens turns those into the 0–100 score and the GO / WAIT / IMPROVE verdict. The model name and version are stamped on every result, so you always know what produced a score.

What Jev is not, here

  • It is not a general chat model asked to “rate this post”. The questions are fixed and closed, and the response is validated against a strict schema.
  • It is not an engagement predictor. Jev scores the writing against the rubric; it has no model of your audience or of reach.
  • It is not a fallback. If the Jev call fails, PostLens surfaces the error — it never substitutes an offline heuristic under Jev's name.

For the request anatomy, read one request, eleven questions; for the factors it scores, the rubric.