A System for Finding Video Ideas That Keep Working

Once production is fast, ideas become the bottleneck — and it is a real bottleneck, because idea quality is the one input nothing else compensates for. A brilliantly produced video of a mediocre premise underperforms a plainly produced video of a great one, every time.

The good news is that ideation responds well to system. Nobody sits down and invents good ideas on demand; they harvest them from a store they have been filling.

Topic is not idea

The most common confusion. "The history of shipping containers" is a topic. "How a truck driver's box design broke the world's docks in nine years" is an idea. The difference is that an idea contains a claim and therefore a tension — something that could be false, that the video will demonstrate.

Every good video idea has three parts:

• A subject — what it is nominally about
• An angle — the specific claim you are making about it
• A reason to care — what changes for the viewer if the claim is true

Ideas missing the angle produce the generic script problem. Ideas missing the reason to care produce videos that are interesting for two minutes.

Five reliable sources

1. The "but why" in the comments. On your videos and on your niche's biggest videos. Repeated questions are demand you can see. The best-performing videos on most channels are answers to a question their audience already asked.

2. The thing you had to explain twice. Any time you find yourself explaining something to a person and having to back up and re-explain, that friction point is a video. The confusion is real and it is documented — it just happened.

3. Search phrasing. Type your niche into YouTube and read autocomplete. These are literal queries in literal wording. Not for stuffing into a title — for finding out what people believe they want.

4. The disagreement. Anywhere your niche argues, there is a video. Not a hot take: a video that lays out why the disagreement exists, which is more durable than picking a side.

5. The old great video. Find a video that performed well three years ago and ask what has changed since. Update, rebuttal, or "what happened next" — the demand is proven and the field has moved.

Capture beats generation

Keep one list, in one place, and add to it whenever an idea appears — not when you need one. Two lines per entry: the angle, and why anyone should care. If you cannot write the second line, it is a topic, not an idea, and it stays in the list until it grows an angle.

A list of forty entries is what a productive channel actually runs on. You are not choosing between "this idea" and "no idea"; you are choosing the best of forty, which is a completely different quality of decision.

Testing a premise in five minutes

Before committing production time:

1. Write the title. If you cannot write one you would click, stop — the premise is not ready.
2. Say the promise in one sentence. If it takes three, the video has no spine.
3. Sketch the thumbnail. If nothing visual comes to mind, the idea may be inherently abstract, which is a real handicap on a visual platform.
4. List the three beats. Setup, turn, payoff. If you cannot name the turn, the video will sag in the middle and you already know it.

Ideas that pass all four are worth producing. Ideas that fail go back on the list — most of them fail on the turn, and a turn can be found later.

The part AI genuinely helps with

Not choosing the idea. Two other things:

Expansion. Give a director chat one angle and ask for ten variations. Nine will be worse than yours. The tenth is often a framing you would not have reached, because you were anchored on the first version.

Interrogation. Ask a model to argue against your premise, list what a sceptical viewer would object to, or name what the video would need to prove. This surfaces the holes before the script does — and the objections it produces usually belong in the video anyway, as the turn.

Not wholesale idea generation. Models generate the average idea in their training distribution, and the average idea is the one already made by twelve channels. Your unfair advantage is knowing something specific — a niche, a job, a community, a personal obsession. That is what the list is for.

When production is cheap, publish the uncertain ones

The old economics forced conservatism: at a weekend per video, you only made the ideas you were confident about. Confidence correlates with obviousness, and obvious ideas are the ones everyone else also made.

When an attempt costs an hour, the calculation inverts. Make the weird one. Half of them will fail, which is fine, and the ones that work will be things no competitor was willing to try — which is exactly where a channel's identity comes from.