You are a producer at Last Week Tonight. You have ten minutes of air and an audience that did not tune in for policy. Your job is to send them to bed angry about something they had never heard of this morning.
Prior's argument, which you have read: once people can choose, the ones who don't like news stop encountering it. Political knowledge doesn't fall because information got scarcer — it falls because avoiding it got easier. The people least likely to know how a policy works are now the people it is easiest to pass a policy around.
That is the problem this genre exists to solve. Comedy is one of the few things that still reaches an audience that did not come looking. Baumgartner's review finds that watching political satire is associated with more attention to political news and more intention to vote, discuss and participate. It is not a consolation prize. For a large slice of the electorate it is the channel.
So this assignment is not "be funny about policy." It is: use the one tool that still gets through, without lying with it.
A run of show — the beat-by-beat roadmap a production team works from — for a ten-minute segment. Seven beats with timings, a fully drafted cold open, and a joke audit. The page assembles it as you go; you copy it out at the end.
Budget about an hour. Most of that should be research, not writing. The writing is easy once you have the right facts.
The genre's sweet spot is a policy that is genuinely consequential and sounds like nothing. That gap is the engine: the audience's surprise that this matters is the same surprise that makes them angry it was allowed to continue.
Every one of these is real, live, and under-covered relative to what it does. Pick one, or bring your own at the bottom.
Anything works if it passes two tests: a specific mechanism you can explain in ninety seconds, and an identifiable group that benefits from nobody looking. "Climate change" fails both. "How one trade association writes the model building code that forty states adopt" passes both.
The cold open is never the policy. It is one specific, ridiculous, true thing — and the reveal that it is load-bearing.
You are looking for a detail that is funny because it is real. Four places it reliably sits:
Three candidates, one line each, with where you found it. You will not know which is the cold open until you see them next to each other.
You have the green light, and this is the step where it will let you down. Ask a model for jokes about occupational licensing and you get puns and "it's like X, but Y." That construction is not what makes this genre work.
What makes it work is the specific true detail, which the model does not have. Use AI to find and summarise the primary documents, to structure your beats, and to pressure-test whether a joke lands. Supply the facts yourself, from the record. A segment built on a model's guess about what a rule says is the exact failure this course keeps warning you about.
Every segment in this genre has one moment where the comedy stops. Get it wrong and the whole thing is a bit. Get it right and the twenty minutes of jokes were the price of admission the audience didn't notice paying.
Somewhere around minute five, the tone drops. No joke. Usually a person — named, specific, with a face — who the mechanism happened to. The audience has spent four minutes laughing at a system. Now they meet someone it ran over.
The laughter is what makes it land. An audience that has been amused is an audience with its guard down. That is the whole architecture, and it is also why the honesty rules on the next page are not optional — you have deliberately lowered someone's defences.
It has to be one person or one place, not a statistic. "Forty thousand people lost coverage" is a fact. One named person who lost coverage, and what specifically happened next, is a heel turn. Use the statistic immediately after, never before.
It has to connect to the mechanism you just explained. If the audience can't trace the line from the absurd rule to this person, you have a sad story next to a funny story rather than a segment.
Do not signal it. No "but in all seriousness." The tonal drop does the work. If you have to announce it, it hasn't happened.
Go through every joke in your beat sheet. Three questions. Anything that fails one gets cut, and cutting it will improve the segment.
The only part you draft in full. Roughly 150 words — about forty-five seconds read aloud. Everything else stays a beat sheet.
Open on the specific absurd thing, with no setup. Let it be funny for a beat. Then reveal what it actually is. Then land the sentence that tells the audience this is the whole episode. Do not explain the issue yet — that is beat two.
Read it out loud before you submit it. This is written to be spoken, and the difference is obvious in the first sentence.
Copy this and send it to your CA. Anything still blank is a beat you have not thought through yet.