Lesson 1: Reading screenshots with vision-language models
19 min read·13 Oct 2026
Module 12 · Lesson 1 · About 55 minutes · Needs: Module 2 Lesson 4 (cost per ticket, the one-cent cap); Module 6 Lesson 1 helps (fixed output shapes)
The situation
It is Monday 17 August 2026, week 12. Customers have started attaching screenshots to their tickets, and the triage model reads only text.
Maya's test case is T-1001 again: Ana Duarte's duplicate charge of 288 USD on invoice INV-2026-004512, now with a screenshot of her Quillpath billing page showing two card payments. Joel asks what a picture costs against his one-cent cap. Maya asks the harder question: when the assistant says what the screenshot shows, how do you know it is right? The screenshot is drawn in code in this lesson; it is simulated, like all Quillpath data.