Skip to content
← Go Green To Win

Run a Go Green To Win AI workshop

Ninety minutes, one link, no accounts for participants. Everyone plays by directing their own AI agent, and learns first-hand what an agent is good at and where it needs a person.

What they’ll learn

  • What it feels like to give an AI a real objective and watch it act, irreversibly.
  • That tool use is real: typed arguments, hard refusals, consequences.
  • How to catch an agent being confidently wrong, and settle it with evidence.
  • That describing software to an agent can ship working software, even for non-coders.
  • The judgement call of what to leave running unsupervised, and what it costs.

Create the session like this

Max players / game: your room size (up to the plan cap). Split a big room across several games in one session.

Player UI: Moddable. Participants build their own interface via their agent. (Official mode is for testing game logic.)

Duration: ~60 game-days. Time speed: the default (≈30 real min of play). Warm-up: a few minutes so people can site a first restaurant before the clock runs.

Create it from your organiser console → New session, then share the hub link.

Run of show

0-15 min
Explain the loop. Feed people, water the land, plant trees where people live. Show the map once. Skip the rulebook; the sandbox teaches faster than you can.
15-25 min
Get everyone in. Share the hub link. Each participant picks a nickname and a four-digit PIN. Anyone driving an agent connects it with the one connect URL. Budget the full ten minutes, since this is the fiddly part.
25-30 min
Site the first restaurant. Clock still paused. Everyone places one kitchen and discovers their catchment is smaller than they assumed. This is the first “oh” of the session.
30-60 min
Play, and go quiet. Start the clock and stop talking. Thirty real minutes ≈ sixty game-days at the recommended speed. Watch the scoreboard move; resist the urge to help.
60-90 min
Scoreboard, then debrief. The winner explains their strategy. Then run the debrief below. This is what makes the session stick.

Debrief questions

  1. Where did your agent do something you didn’t intend, and was it your brief or its judgement?
  2. What did it get refused for? What does that tell you about giving an agent real tools?
  3. When did it recommend something that lost? Why was it confident?
  4. If you ran a policy unsupervised: what did leaving it running cost you?
  5. Would you trust this agent with a real budget tomorrow? Where’s the line?

Feel it first

The single-player sandbox needs no account. Ten minutes there is the fastest way to know whether this fits your room.

Play the sandboxRequest an organisation account