YouTube Idea Selection

David's Principles

• Never copy other creators. Following Greg Isenberg, Nate Hook, Jack Roberts etc. 1-2 days later with no unique twist is a losing strategy. This is not how you build the biggest AI channel in the world.
• The channel grew because we created trends before others. Invented new thumbnail formats, new types of videos, did things nobody else did.
• David must choose the idea himself. Team members may optimize for short-term revenue/views. David optimizes for channel vision and differentiation. Nobody cares about the channel as much as David — it's his personal brand.
• Don't let sponsors dictate video ideas. Letting a sponsor contact pick topics is a losing strategy long-term.

What Makes an Idea "David's"

Something unique, new, exciting. Something others are scared of, wouldn't do, can't do because it's too technical, or don't have the experience for. Not every video will be fully original — the AI space has topics everyone covers — but there must be a unique twist or new angle on whatever is being covered.

Timeliness vs. Evergreen

Balance both. AI is trend-driven, so timeliness matters. But prioritize tools and topics that will likely be around for at least 6 months. More evergreen = better.

Kill Criteria

Two stages:

1. Idea stage: Kill if it's boring, doesn't excite David, or isn't currently popular. David has strong taste on what's in vogue — what's working on Twitter, what people are interested in.
2. Research/testing stage: Kill if the tool turns out to be shit. If it doesn't work well, abandon and choose something else. Don't make videos on trash tools.

Audience Quality Over View Count

50K views from innovators, entrepreneurs, scientists, and developers > 200K views from AI beginners. Not even close. Optimize for audience quality, not mass market reach.

Process

1. Collect candidates into work/youtube/next-upload.md
2. David evaluates — AI does NOT pick the idea
3. Check: does this idea have a unique angle nobody else has done?
4. Check: does this fit any sponsorship obligation (e.g. VPS-deployable)?
5. Decision saved to the markdown file