Lesson 1: Why long runs go wrong: context is scarce
32 min read·13 Oct 2026
Module 7 · Lesson 1 · About 45 minutes · Needs: Module 5, Lesson 1 (run_agent,ScriptedModel) and the idea of a token (Module 1)
The situation
It is Tuesday 26 May 2026, week 7. Dev has an idea: "Run the morning's ten requests through one conversation, a shift. Then the agent can see that R-1006 is Aisha repeating R-1001." You try it on the test queue.
It works, and every call gets bigger. Priya asks: "By the afternoon, will it still remember that Aisha's $129.00 refund is waiting for Marta?" This lesson traces that one fact through the shift.