AI workshop · 10 September 2026

Choose a track.
Start together.

One starting point for two workshop tracks. You will work as a team, review the AI's proposals, and finish with something you can show and explain.

Two tracks Teamwork Evidence before claims

01 / Choose a track

Where will your team work?

Choose the track shown on your team assignment.

Developer teams · 60 minutes

AI mob programming

Find and resolve one contract inconsistency. Four people share one browser workspace and one AI agent, then defend the result with evidence.

One inconsistency in a C++ or C# wrapper or test
A smart primary agent with lower-cost subagents
Rotating roles and a timed assignment
Tests, diffs, and evidence before approval

For non-technical functions

Northstar Rescue

Choose a role and make sense of a messy, synthetic business case. Use AI to find gaps and prepare a decision — without guessing.

Sales, Marketing, Finance, HR, or Leadership
Conflicting sources and realistic trade-offs
One concrete recommendation to present

02 / Developer track

Your 60-minute assignment

One laptop, one shared workspace, four roles rotating every 15 minutes.

Your mission: find and resolve one contract inconsistency in your C++ or C# workspace. The specification wins over code, tests, and AI output.

The inconsistency may be in the wrapper or a public test. Do not assume which: state a hypothesis and design evidence that can distinguish the alternatives.

DriverOperates the shared workspace NavigatorFrames the next agent request Test leadOwns checks and evidence Domain scepticChallenges claims against the spec

0–12Orient and establish a baselineRead the specification, trace the wrapper, and run your language checks.
12–42Diagnose before editingState a hypothesis, design a discriminating test, then make the smallest justified change.
42–55Validate and documentRun all checks, submit with workshop-check, create a solution commit, and record its hash.
55–60Share one findingExplain the contract, the evidence that changed your mind, and one uncertainty.

Done when

The diff is understood, the checks are recorded, and your team can defend the result against docs/spec.md.

1 / Open your workspace

Team 01 Open ↗ Team 02 Open ↗ Team 03 Open ↗ Team 04 Open ↗ Team 05 Open ↗ Team 06 Open ↗ Team 07 Open ↗

2 / Connect the team key

Paste your team's OpenRouter key into the hidden prompt. OpenCode starts automatically. If the credit runs out, click OpenRouter key in the status bar and paste the replacement.

3 / Start the agent together

Read AGENTS.md and ASSIGNMENT.md, then the specification, relevant source, and public tests. Do not edit yet. Delegate the initial repository map to the explore subagent, then summarise our current phase and first verification command.

Shared safety rule

Use only the material provided for the exercise.

All workshop material is synthetic. Do not add customer data, Carmenta private source code, passwords, credentials, or private messages.

Choose a track.
Start together.

Where will your team work?

AI mob programming

Northstar Rescue

Your 60-minute assignment

1 / Open your workspace

Choose your team's perspective

Should we pursue the opportunity — and how?

Use only the material provided for the exercise.

Choose a track.Start together.

Where will your team work?

AI mob programming

Northstar Rescue

Your 60-minute assignment

1 / Open your workspace

Choose your team's perspective

Should we pursue the opportunity — and how?

Use only the material provided for the exercise.

Choose a track.
Start together.