Where we are.
Episode 1 runs in eight parts, about 29 minutes, then questions. On air the yellow card is the part we are in. Everything you see is in the public repo, so you can follow along or come back to it later.
1The argumentThe claim in one line, and the five things "establish the run" can mean. The plays can test two of them; we rule on one.
2The playsA quarter of a million NFL plays, 372 columns each, cut to the 18 the question needs. The betting lines are not among them.
3The tool, by handOne command, one number that looks like proof, and the list of filters that shows why it is the wrong number.
4The policy, built from an outlineEight bare headings filled in on air: the written job description that turns a model into an analyst you can check.
5Run itThe runner in five beats, then the episode's claim goes in: five tool calls on screen, one ruling out.
6Claims from the roomYour claim from the chat, unrehearsed. Then a claim the agent has to decline, and the reason it gives.
7CloseWhat carries to your own agent: a policy file, one tool, an allowlist, and filters you can see.
8The season sheets, then Q&AThe five-season answer with its evidence, this season's sheet made live, then your questions.
Space next
Backspace back
R start over
While you wait
The episode page explains the setup for someone who has never built an agent. The plays can be read row by row, the explorer runs the tool's real output under every filter, and the tool page is a diagram of the one script the agent is allowed to run.