Blueprint — layout C
A crew of named AI agents, one shared stream, five artifacts — here's what's actually running underneath.
Join your group →The architecture behind the Execathon
Each participant's own chat pane — per person, not synced across your team.
The one piece of state every teammate shares — market snapshot, product concept, campaign, hero image, web page, updating live for everyone in the group.
Named agents, addressed directly by @name. Every request is attributed to your group specifically.
Text generation calls the Anthropic API directly — per-group attribution on every call, no shared pool.
Hero images call the OpenAI API directly, through the same per-group attribution layer.
How it works, and how you're scored
Your group is its own isolated workspace — nothing you see or send is shared with any other group. Inside it, you address a crew of named agents directly, in one shared stream everyone at your table can see.
@analyst, @pm, @marketer, @designer — addressed by name, no tabs, no dropdown. Anyone at the table can talk to anyone on the crew.
A market snapshot, a product concept, a campaign, a hero image, and a web page — five shared slots, synced live to your whole team.
Your conversation with the crew is your own — not synced across teammates. The five artifact slots are the shared state everyone sees update together.
Process-based — how well your team operates the cycle: briefing clearly, iterating, using the hour well. Not a simulated audience reacting to what you ship.
What this is
Every team gets the same problem statement at the same moment. There are no personas, no pre-built templates — just your team, a crew of agents, and the hour.