My most expensive creative experiment
and why it changed my strategy.

You care about 3 things: Cost. Speed. Execution quality.

So here is the real question: can AI meaningfully reduce production cost, accelerate creative execution, and still produce assets good enough to drive revenue?

The Experiment

To test this, I took standard Alfa Romeo website photos and built an Image-to-Video workflow to replace traditional motion production for certain assets.

THE STACK

What actually worked

The snowy driving sequence at 0:14 proves the concept. Energy. Atmosphere. Depth. It is absolutely good enough for performance environments and early creative testing.

The Reality Check (Honest failures)

Even with engineering discipline, today’s models still break in real-world conditions.

• Hand shifting stick is unstable
• Wheel caps deform during motion
• Interior navigation display flips to rear zoom
• Driver face inconsistencies
• Micro-physics still drift

Could I brute force fix it by burning more compute and time? Probably. Did I? No.

Because in two months, newer models will likely solve most of this out of the box. Which means the smarter move is building systems, not over-optimizing today’s weaknesses.

"Creative scale is not about AI output. It is about systemized content creation. AI is just the execution layer."

The System Strategy

This experiment is not about "AI making cool car videos". It is about building repeatable creative infrastructure. I treated it as engineering, not creativity chaos.

I used variables to maintain output control. Below is the exact checklist I use to prompt the AI.

Why this matters for growth

• Lower dependency on production
• Cheaper creative testing
• Faster cycles = more learning
• More learning = better performance

If growth compounds through execution quality and iteration speed, this is infrastructure.