You save between five and twenty things a day. A briefing is ten minutes long. So the question is never "how do I summarise all of this", it is "what deserves airtime, and how much".
A radio programme answers that before the writing, in the running order. We do the same. Before a single sentence of the script is drafted, every selected story gets a duration in seconds, and every dropped story gets a reason.
The pilot episode's running order
| Segment | Duration |
|---|---|
| Intro — today's menu | 30 s |
| Private credit, the cockroach of Wall Street | 150 s |
| An AI placing stock orders | 130 s |
| The twelve possible endings of AI | 180 s |
| A photography exhibition | 90 s |
| Outro — sources recap | 20 s |
Those numbers are not decoration. A hundred and fifty seconds is roughly three hundred and thirty words of documentary narration. The writer receives that constraint at the same time as the sources, and writes to the format. A ninety-second story has no room for historical context: it opens on the fact, gives two figures, and gets out.
Why not afterwards?
We tried the other order, writing freely and then cutting in the edit. The result is bad, for a mechanical reason: when you cut a finished text, you remove the sentences that are easiest to remove. Those are almost always the specifics, the figures, the caveats. What survives is the general scaffolding, which is the least informative part.
Deciding the duration upfront does the opposite. The constraint forces a hierarchy inside the sentence: what do you keep when you only have three hundred words? You keep what is new and verifiable, and you drop the context the listener already has.
Dropped stories are content, not waste
A running order keeps four stories and often drops as many. Each exclusion carries an explicit written reason: already covered in a previous episode, single weak source, redundant with another chapter, too thin to fill sixty seconds.
Those reasons are not thrown away. They surface in the app, and you can see why the link you saved did not air this morning. That is what makes the selection arguable, and therefore credible. A machine that drops things without saying why looks like a machine that lost your links.
The rule that follows from all this
The system also remembers what it has already explained. A concept covered on Monday is not covered again on Thursday: it is mentioned, and the time saved goes to what has changed since. On a daily product, repetition is the first enemy. It loses listeners well before imprecision does.
Ten minutes is short. Deciding in advance how they are divided is what separates a programme from a list of summaries stacked end to end.