Jev Academy

Lessons

The full curriculum has 12 steps. The home checklist follows the shorter path, steps 0 to 7.

  1. 1

    What System One is

    Jev is built for software, not chat. It answers typed questions with structured answers and probabilities.

  2. 2

    When to use Jev, an LLM, or plain code

    Code runs the workflow. Jev makes quick judgment calls. LLMs do the writing. Keep fixed-rule work in plain code.

  3. 3

    Set up and check your API key

    Make a dashboard key, store it as TYPESAFE_API_KEY, and check it works without ever printing it.

  4. 3

    Make your first API call

    Send state plus one Noul question to POST /v1/systemone. Examples in curl, JavaScript, and Python.

  5. 4

    State: what gets judged

    State is the material the model judges. Send it as a string, an object, or an array. Only include what the questions need. Keep facts in the state and judgments in the questions.

  6. 5

    Noul: yes or no questions

    Noul answers a yes/no question with a number from 0 to 1. Higher means more likely yes. Optional criteria fields let you define yes and no. There is no confidence field.

  7. 6

    Choice: pick one from the menu

    Choice picks one option from a fixed list. You get the pick, per-option probabilities, and a confidence. Add other or none when the list might not cover every input.

  8. 7

    Score: rate it with a rubric

    Score rates something against an ordered rubric, like star ratings. The result can land between levels. Don't use it for exact counts, math, or dates.

  9. 8

    Confidence: how sure before you act

    Every answer tells you what. Confidence tells you whether to act on it. High means go, medium means check, low means hand it to a human.

  10. 9

    Fan-out: many questions, one call

    Ask many questions in one call, even ones you might not need. Your code picks which answers matter. It's cheaper and faster than one call per question.

  11. 10

    Combine patterns in code

    Four patterns to mix and match: intent routing, composite scoring, confidence-gated routing, and fan-out. Instructions and criteria can also take JSON.

  12. 11

    SDK labs: JavaScript and Python

    Install the JavaScript or Python SDK. Send one document plus a Choice, a Noul, and a Score in a single call. Then pick a model.

  13. 12

    Capstone labs: practice only

    Four practice labs: request routing, newsletter triage, app intake, and Fit Desk adherence. Learning only. Do not touch the real product repos.

Path steps 0 · 1 · 2 · 3 · 4 · 5 · 6 · 7 — see home.