Season 2 · Prompt Engineering · Episode 6

S2E6 The Council

Five advisors who argue with each other so the blind spots show

What this lesson is about

Ask an AI "is my idea good?" and it tends to say yes. Ask "what's wrong with my idea?" and it finds problems. Same idea, opposite answers, because the response reflects how you framed the question. The Council is the structural fix. Instead of asking one question from one angle, you convene five advisors who each think from a completely different perspective: a Critic, a Big Picture Thinker, an Optimist, a Fresh Eyes, and a Doer. A Chairman then synthesises where they agree, where they clash, and, most usefully, what all five of them missed.

What you'll be able to do

Bring a real question, a decision you are genuinely unsure about with more than one reasonable answer. Good council questions have real stakes and genuine uncertainty, like "Should I choose Higher Physics or Higher Biology?" Questions with a single right answer, or tasks like "write me an introduction," do not work as well. The idea is inspired by Andrej Karpathy's LLM Council, where several AI models peer-review each other anonymously.

If the answer changes when the question changes, which answer was ever true?

Readings

Where these facts come from

Every source on this page was opened and checked before it went here. Click through and check anything you are about to rely on.

Key words

Framing
How a question is put. Ask what is good and you get strengths; ask what is wrong and you get faults.
Sycophancy
The pull toward telling you what you want to hear — the thing the Council is built to defeat.
Synthesis
The Chairman's job: where the five agree, where they clash, and what all of them missed.