Guide

How to decide what to watch.

A method you can run by hand in a few minutes, and the one Worthwatch runs for you. It rests on the transcript and three questions.

The title and the thumbnail are an offer.

A title is written to be clicked. It names the biggest promise the video can carry. The thumbnail adds a face and three words.

Neither tells you where the substance starts, how much of the runtime it fills, or whether the promise is kept. The transcript tells you all three.

Step one: open the transcript.

Under the video's description there is a "Show transcript" button. It opens every caption line with its time. Auto-generated captions are rougher than human ones, and still enough for this.

Scroll it once, fast, before you read a line. You are looking for the shape: where the talk about the topic begins, where it stops, and what sits in between.

Step two: ask three questions.

  • Where does the substance start?

    Find the first line that is about the topic rather than about the channel, the sponsor, or what is coming. Note the time. That is where you press play, if you press play.

  • Does it deliver the title?

    Search the transcript for the key word in the title. If the word appears once, near the end, with "link in the description", the title is not kept. If it appears throughout with specifics, it is.

  • What can be skipped?

    Sponsor reads announce themselves. Repeated points repeat their phrasing. Filler is the stretch where nothing new is said. Mark the ranges.

Step three: say the verdict out loud.

Watch, when the substance starts early, fills most of the runtime, and keeps the promise.

Skim, when there is substance but it sits in parts: watch the parts.

Skip, when the title is not kept, or when the video says nothing beyond its title.

Say it in one sentence with the reason, as if to a friend. "Skip, the fix is a link and the rest is the story of the problem." If you cannot say the sentence, you do not have a verdict yet.

Habits that make it quick.

Judge the video against its own promise. The best video on the subject is a different question, and a modest title that is kept is a watch.

Never judge the creator. The verdict is about this video for you, today. A channel you skip five times may post the one you need tomorrow.

Keep the sentence. A verdict without a reason is a mood.

What Worthwatch does with the same three questions.

It reads the same transcript, in your browser. It cuts it into 30-second blocks and asks a model the three questions with a strict form for the answers. Code turns the model's block numbers into timestamps, so every time in the panel is where the words are.

It shows the verdict with its sentence, the start of the substance, the points with their times, the skip ranges, and the promise check. It says, on every video, that the verdict is judged from the transcript and can be wrong.