Prompting & Contextintermediate11 min · 9 steps
Chain-of-Thought
Intermediate reasoning steps make hard problems tractable. Reasoning models now do this internally, so you control them with effort, not with prompts.
requesterrorresponsetokentool callheavy work
ProblemStep 1 of 9
One forward pass, one guess
An LLM predicts the next token. For a multi-step word problem, the first answer token has to carry all the arithmetic at once, with nothing written down in between.
For anything beyond a lookup, a single pass often isn't enough, and the model produces a confident wrong number.
Lesson 3 of 8 in Prompting & Context