Module 5 — Due before Module 6

Your Assignments

Two assignments this week. One is about developing a verification habit. The other is about starting to think seriously about the group project.

Assignment 1 Verification Practice

Cross-Model Validation on Real Work

Take one piece of AI-generated output from the past two weeks that you actually used, or plan to use, in a real decision or deliverable. Something where being wrong would matter.

  1. Run the same underlying question or task through a second platform. If you generated the original in Claude, use Perplexity for anything requiring current information, or ChatGPT for a second structural take.
  2. Note where the outputs agree. Note where they diverge. Note where one platform added something the other missed entirely.
  3. Decide whether the divergence changed your confidence in the original output, and in what direction.
Bring to Module 6

What you validated, what surprised you about the comparison, and whether the exercise changed your assessment of the original output.

Assignment 2 Project Thinking

Name One Problem Worth Solving Together

Modules 7 and 8 are collaborative builds. Your cohort will select a shared challenge, design an AI-assisted approach, and present at the combined capstone in late April.

You are not selecting the problem today. You are thinking about it so the Module 6 selection conversation has somewhere to go.

  1. Identify one problem in your industry, organization, or function that you think is worth a group's attention. Not a safe, generic problem. Something that would actually be useful to work on.
  2. Consider whether it can be addressed in two sessions without requiring you to share proprietary data in a group setting. The capstone is cross-cohort. Abstract the problem so it can be discussed openly.
  3. Write two or three sentences describing the problem in terms anyone in the group could understand, not just someone with your specific context.
Bring to Module 6

Your two-to-three sentence problem statement. You will have a chance to pitch it and hear what others bring. The group decides together.

Multi-Model Workflows

Module 6 moves from understanding the landscape to building within it. Before the session, spend 15 minutes with the Multi-Agent Patterns reference. Pay particular attention to the Creator-Critic pattern and the Relay Race framework. Those are the two we will practice live.

Required — 15 min
Multi-Agent Patterns
Five workflow patterns for using multiple AI models in sequence. Includes de-hype calibrations and practical entry points for insurance use cases.
Optional — 10 min
Agentic Email Processing
An applied example of multi-model workflow design. Human control points, config governance, and what "agentic" actually means in practice.
Optional — 12 min
Why Multi-Model Consensus Catches Hallucinations That Single Models Miss
A practical case for the verification practice in Assignment 1. Thor Matthiasson builds the argument around a fabricated academic citation as the test case. Read the limitations section — the part on shared confabulation is the most important part.
Fri April 10 Vinnie Garth Field Trip — Cross-cohort lunch-and-learn. Vinnie walks all three cohorts through his ChatGPT-native executive operating system. 11:30 AM to 12:30 PM ET. Details to follow.
Module 6 Multi-Model Workflows. Group project problem selection begins. Come with your problem statement ready.
Late April Module 9 combined capstone. All three cohorts. Each cohort presents their group project. 90 minutes.