How to stop wasting hours on YouTube tutorials that go nowhere

The honest number, across everything we have measured: 20% of videos are not worth the time they ask for, and the median video that is worth it still spends 40% of its runtime restating itself. The waste is not a handful of bad videos. It is baked into the format.

That means the fix cannot be “pick better videos” — you are already trying to do that, and the format is working against you. What works is running the same short filter before you press play, every time, instead of trusting the thumbnail.

The problem is not attention, it is information you do not have yet

Nobody sits down intending to watch a bad tutorial. The title looked specific, the thumbnail looked competent, the view count looked respectable. All three of those signals correlate weakly at best with whether the video delivers — a title can promise a mechanism and deliver an anecdote, and views measure whether people clicked, not whether they got what they came for. The information that would actually help — how much of the runtime is setup versus substance — does not exist anywhere on the page. You find out by spending the time.

This is why “just be more selective” fails as advice. Being selective requires a signal to select on, and the platform hands you three weak ones and calls it enough. The four checks below are an attempt to replace those weak signals with slightly stronger ones — still proxies, still imperfect, but each aimed at something closer to the thing you actually care about than a thumbnail is.

Check one: does the title name a specific thing, or a feeling?

A title that names a mechanism, a number, or a concrete step is a different kind of video from one built around a promise. “How connection pooling works in Rails” names a thing. “The one trick nobody tells you about databases” names a feeling. The first is falsifiable — you can check afterward whether it delivered what it said. The second cannot fail, because it never specified anything to fail at. This single check filters out a meaningful share of the videos that later turn out to be padding.

Check two: read the first thirty seconds for specificity, not interest

Play the video and stop after thirty seconds, before any commitment. The question is not whether the opening is engaging — most openings are, that is what they are built for. The question is whether it says something specific in that time. A video that opens with the exact problem it will solve tends to keep behaving that way. A video that opens by telling you this trips people up constantly has used its most valuable thirty seconds to say nothing measurable.

Check three: ask whether the idea compresses to a sentence

If someone asked you to write the video’s core idea as one sentence, could you, without having watched it? If the honest answer is yes and the runtime is twenty minutes, the other nineteen are structure around one idea — introduction, restatement, recap, outro. If the idea genuinely does not compress — each step depends on the one before it — a long runtime is more likely doing real work. This is the check most people skip, and it is the one that predicts the outcome best.

It also explains a pattern that otherwise looks random: two tutorials with the same runtime, in different subjects, can have completely different amounts of real content in them. A twenty-minute video on a method that fits in one sentence and a twenty-minute video on a process with a dozen dependent steps are not comparable just because they share a runtime — the second one has somewhere for twenty minutes to actually go.

Check four: look at where the comments cluster

Not the sentiment, the location. A comment section where multiple people have independently posted the same timestamp is a crowd telling you, for free, where the payoff actually sits relative to the runtime. If that timestamp is close to the start, the setup was short. If it is close to the end, or if nobody has posted one at all, treat that as a small negative signal rather than a neutral absence.

Running all four takes less time than the video you were about to open

None of the four checks requires watching past the first half-minute, and together they take under a minute. That is the entire point: the checks are cheap because they are designed to be run before the cost, not instead of it. Skip any one of them and you are back to trusting the thumbnail.

A useful way to think about the order: each check is cheaper to run than the one before it is to skip. Reading a title costs nothing. Watching thirty seconds costs thirty seconds. Both are trivial next to the twenty minutes you save by not opening a video that fails either one. The asymmetry is the entire argument for running the filter at all — it is not that the filter is perfect, it is that even an imperfect filter is dramatically cheaper than the thing it is filtering.

What the checks cannot see

All four are proxies read from outside the video. They predict density reasonably well and predict correctness not at all — a confident, well-titled, specifically-opened video about something untrue passes every check here. That limit is real and worth stating rather than hiding: judging density and judging truth are different jobs, and this filter only does the first one.

There is a second limit worth naming too: the checks predict whether a video is dense for its subject, not whether that subject is the one you actually needed. A sharp, efficient, well-made tutorial on the wrong tool for your problem still wastes the time it takes to watch. No filter run before pressing play substitutes for knowing what you are trying to solve in the first place.

Automating the parts that can be automated

Two of the four checks — how much of the runtime is restatement, and where the useful part actually starts — are exactly the numbers Gistil measures and puts on the YouTube page itself, before you press play. The other two, reading the title for specificity and reading the comments for a cluster, are still yours to run; they take seconds and no measurement replaces judgment about what you are actually trying to learn.