Lesson 2 · Five Core Concepts

🎯 What You'll Learn

• Master the five fundamental terms in Jev's architecture
• Distinguish the three primitive question types: Noul, Choice, Score
• Grasp the genuine definition of confidence

Terminology Reference Table

A True Decision Engine

With ChatGPT, you usually prompt for a long response and use regex to parse out "refund". Jev reverses this paradigm: you declare the valid answer space upfront, and Jev selects the outcome with mathematical certainty.

The Three-Part Pipeline

• State: The raw contextual evidence or customer payload.
• Questions: The schema of questions you define (Noul, Choice, or Score).
• Answers: Keyed structured results matching your question identifiers.

Three Primitives Cover Everything

• Noul: Boolean judgment returning a probability between 0 and 1.
• Choice: Single-choice classification among explicit options.
• Score: Ranked ordinal scoring according to criteria.

Confidence Means Distribution Concentration

A confidence of 0.92 simply means probability is sharply focused on this outcome. Always validate against historical enterprise samples to set an automated routing threshold.

✅ Quick Quiz

1. [Multiple Choice] What are the three parts of an invocation? A. Prompt, Response, Token B. State, Questions, Answers C. Input, Model, Output 2. [True/False] Confidence of 0.92 guarantees 92% real-world accuracy. 3. [Multiple Choice] To determine if a message is urgent, which primitive should you pick? A. Noul B. Choice C. Score

Answers

1. B. State, Questions, and Answers. 2. False. It reflects mathematical concentration; benchmark with real samples. 3. A. Noul returns a 0–1 probability for boolean conditions.

📌 Takeaway

State + Questions + Answers: three primitives build any decision pipeline.