On 30 September, flydubai flight FZ1073 from Dubai to Tel Aviv dropped about 17,000 feet in a minute and a half. According to Israeli officials and passengers, the first officer had attacked the captain. The injured captain, Smit Machchhar from Mumbai, kept fighting, passengers restrained the attacker, a plumber pulled the nose up because he had watched Air Crash Investigation, and an off-duty pilot landed the plane safely.
Within a few hours of starting, we had a seven-and-a-half-minute animated documentary on it, at the production level of the best explainer channels on YouTube:
iNormally that is one to two weeks of work for a team: research, script, illustration, animation, voice, music, sound, editing. My hands-on time was about two hours. For most of it, I was eating bun maska.
What I did
- Wrote a messy brief: the plane must be animated, and no cartoon look, because this is serious.
- Answered a few questions Claude Opus 5.5 asked: real faces for the people involved, but no name and no face for the accused.
- Added one idea: warm colours while the flight is calm, the whole screen turning red during the attack, warm again on landing.
- Reviewed the cut and asked for more. Version two came back in the same sitting.
That’s it. I decided and reviewed. The AI did everything else.
What the AI did
It fact-checked me. It built a sheet of over a hundred facts, each marked verified, claimed or disputed, and two of my assumptions failed. The attacker’s nationality was only a claim by Israeli officials, so the film attributes it. “Two British pilots” turned out to be one passenger’s account, so the film says exactly that.
It checked itself. The narration was transcribed back to catch mispronounced names, and the captain’s surname was fixed. A background that looked like blood under the red colour was caught and replaced.
It built in parallel. One chapter was made first as the quality bar, then several AI builders made the rest at the same time.
The bill: about a dollar of images, some voice and music credits, and free rendering on a laptop.
Three lessons for managers
1. Your job becomes directing, not doing. About ten decisions mattered more than any hour of execution.
2. Make your AI show its sources. The most valuable step was the research pushing back on me.
3. Your taste is the new bottleneck. When execution takes hours, what matters is knowing what good looks like.
From first idea to publish button, with a bun maska in hand. Reply and tell me what you’d make with it.
Until next time.


