CourseLarge Language Models: The Complete Course · Module 3: Inference and Decoding · part 24 of 103
Part 24 · Module 3: Inference and Decoding

Lesson 3: Forcing a format, and paying for thinking

28 min read·13 Oct 2026
Module 3 · Lesson 3 · About 50 minutes · Needs: Module 1 Lesson 1 (the setup cell and the tickets) and Module 2 Lesson 4 (what a ticket costs); Lessons 1 and 2 of this module help

The situation

It is Friday 19 June 2026, week 3. Two questions arrive about T-1001, Ana Duarte's duplicate charge of 288 USD.

Tomasz's queue tool reads your triage label as JSON. Yesterday a reply began "The category is", and the parser stopped. He asks: "Can the model be made unable to write anything else?"

Joel has the usage log open. The course model, gpt-oss-20b, writes hidden reasoning before every label, and he pays for it. He asks: "How much of that thinking do we need?"

This lesson answers both, on T-1001 and then on the six refund tickets.

The rest of this course is yours to keep

This course is bought on its own, once, and stays readable afterwards, including the parts added to it later.