Week 51: Using AI to validate project ideas

(Before you invest a year on them)

I have a theory: distribution from 2026 is going to be even more about people, and less about SEO. The cult of personality, as much as I find it a bit cringe, isn’t going anywhere. YouTube is massive and growing. Short-form video is where attention lives.

My new idea: 365 Great Records. One vinyl album a day, 30-45 seconds each, posted to TikTok and YouTube Shorts. I’d batch-film 10-14 at a time, change shirts between takes to fake different days, and use Descript to edit fast.

But here’s the thing: 365 is a nice number. It’s also a full year of commitment. Before I signed up for that, I wanted to know if the workflow was even sustainable. Could AI help me prototype this thing fast enough to make a real decision?

The Process

Here’s how I stress-tested the idea:

Structuring the concept with AI I used Claude to figure out what this project should actually look like. Platform strategy (YouTube Shorts + TikTok, manual uploads to both). Format options. How to vary my angle on each album so I’m not saying the same thing 365 times. AI gave me a framework: weird comparisons, personal stories, backstories, technical details, emotional impact. Different lenses for different records.

Scripting the first batch I asked AI to help me script the first 7 albums. This forced me to articulate what I wanted to say about some of the records I bought this year. Having scripts meant I wasn’t fumbling on camera.

The physical problem AI couldn’t solve Here’s where it got real. AI could tell me to use a ring light, prop my phone at eye level, film in 1080p/30fps vertical. But actually framing the shot? Figuring out where I should stand, how to hold the record, what looks good on camera? That was trial and error. I had to look at myself and be judgemental about it. No prompt fixes that.

Editing in Descript Once I had the ring light set up and figured out framing, filming was quick. The real win was Descript: upload the footage, edit by transcript, cut dead air and ums automatically, add overlays, stitch clips together, export. The whole thing felt surprisingly lean.

Timing the workflow About 25-30 minutes per finished video, once I had the setup dialed. That would probably get faster with practice. Batch filming would help. But even at 15 minutes, that’s 90+ hours across the year, not counting the time picking albums, writing scripts, and posting.

The Outcome

The workflow works. AI helped me get from vague idea to filmed prototype in a day, not weeks. Specifically:

AI structured the concept and gave me format variations
AI helped script the first batch so I had something to say
Descript made editing genuinely fast
Total time per video: 25-30 minutes (with room to improve)

But I haven’t posted anything. And I haven’t decided if I’m going to continue.

The reason isn’t the workflow. It’s the opportunity cost. A year is a long time. Every day I spend on 365 Great Records is a day I’m not spending on something else. And when I asked myself “is this fun?” the honest answer was: I don’t know yet. Doing something 365 times is different from doing it 7 times.

Key Takeaway

AI can dramatically reduce the friction of making content. It can help you structure ideas, script content, plan workflows, and edit fast. What it can’t do is tell you whether the thing is worth making in the first place.

The real value here wasn’t producing 7 short videos. It was getting to the real question fast enough that I hadn’t already sunk months into the project. AI turned a potential year-long commitment into a week-long test drive.

Pro Tips

Use AI to prototype before you commit: Don’t spend weeks setting up infrastructure for an idea you haven’t validated. Use AI to get a rough version out fast, then decide if it’s worth continuing.
Separate workflow questions from motivation questions: AI can answer “how do I do this efficiently?” It can’t answer “do I actually want to do this 365 times?”
Accept that some problems aren’t promptable: Framing a shot, finding your on-camera presence, deciding if something is fun—these require you to do the thing and pay attention to how it feels.
Descript is genuinely good: If you’re making any kind of video content, editing by transcript and automatic um-removal is a massive time saver. Worth the subscription.

Want to Try It Yourself?

Descript for fast video editing (edit by transcript, auto-remove filler words)
Claude or ChatGPT for structuring concepts, scripting, and planning workflows
Ring light + phone tripod for consistent, decent-looking footage
A project you’re curious about but haven’t committed to — the best use of AI might be killing bad ideas early