Strategic AI Architects
Free Audit Join the Community

Build a Video to Shorts Pipeline

Step 7 of 9 · About 30 minutes per episode

Build the Open Loop Cold Open

The longform episode needs a first fifteen seconds that opens a loop and refuses to close it. This is the difference between an episode people watch and a recording people scrub.

What you get

A cold open cut, an episode title, and the timestamp where the loop closes later in the episode.

The shorts are done being the focus for a moment. This step is about the longform episode, and it is the one thing that most determines whether anyone watches past the first minute.

What an open loop actually is

You take the strongest reveal in the whole recording and you play it first, then cut it off right before the payoff. Mid sentence is ideal. Something like and that is when I realised the whole thing costs, hard cut, title card, episode starts from the top.

The moment stays in place later in the episode. When viewers reach it, the loop closes and they get the payoff they were promised. That is the contract. Break it and they stop trusting your cold opens.

  1. The tease must be coherent up to the cut. Same rules as the coherence gate. A tease that is confusing is not suspenseful, it is just confusing.
  2. The payoff must actually exist later. If you tease something that never lands, you have made a clickbait episode.
  3. Cut before the number, not after. The specific detail is the payoff. Give away the setup, withhold the figure.

Build it

Paste into Claude Code
Create skills/openmic-intro/SKILL.md, then run it.

The skill takes the word timestamped transcript and the rough cut EDL, and:

1. Scans the FULL transcript for hook candidates: bold claims, surprising
   numbers, live demo payoffs, emotional peaks. Ranks the top 5 with a reason
   for each.

2. Builds a suspenseful open loop cold open of 10 to 25 seconds. It picks the
   strongest reveal and CUTS IT OFF right before the payoff. Mid sentence is
   ideal. The tease must be coherent up to the cut point, and the payoff must
   actually exist later in the episode.

3. Emits: the tease segment timestamps, the exact cut point, an episode title
   of 60 characters or fewer built on curiosity plus specificity, a one line
   premise, and the timestamp where the payoff lands.

4. Specifies the sequence: tease, hard cut, 2 second title card, then the
   episode from the top. The teased moment STAYS in place later so the loop
   closes when viewers reach it.

Fail condition the skill must respect: if no candidate is a genuine open loop
tease, say so and recommend recording a 15 second custom intro instead of
forcing a weak one.

Now run it against work/ep-test.words.json and show me the top 5 candidates
with your reasoning before you commit to one.
Check it worked
Watch only the first fifteen seconds. If you already know the answer, you cut too late. If you have no idea what the episode is about, you cut too early or picked the wrong moment. The target is knowing exactly what question is about to be answered and not knowing the answer.
Where this bites
Do not force this. Some recordings genuinely have no reveal worth teasing, usually the ones that were mostly logistics. Recording a fifteen second custom intro is a better episode than a cold open built from a moment that was not actually surprising.

Questions

Should the teased moment be removed from where it originally happened?

No. Leave it in place. The whole mechanism depends on the viewer reaching it later and getting the payoff. Removing it means the loop never closes.

How long should the cold open be?

Ten to twenty five seconds. Under ten and there is not enough setup for the cut to sting. Over twenty five and you have made a short instead of an intro.

Talk it through

Stuck on this step?

Free 30 minutes. Bring the error, the transcript, or the clip that will not pass the gate, and we will work through it.