In this section I review one AI-powered application and demonstrate how it can be used to create new value.
At 11 PM a few nights ago I was watching the Artemis II crew live as they rounded the far side of the Moon. As a kid I wanted to be an astronaut, so this was a genuinely magical thing to watch. Then, on live TV, a jar of Nutella floated across the Orion capsule.
I forwarded the clip to my family WhatsApp. My sister reminded me of the 2013 heist, when 5 tons of Nutella vanished from a warehouse in Germany. A fun idea came to me. What if the astronauts were the thieves? Trained thirteen years for this moment, launched to the Moon, and brought the last jar with them.
By midnight I was producing a 60-second fake conspiracy ad. Claude Cowork was my creative director, writing the script, breaking down scenes, producing prompts for every shot, pulling clips from real NASA footage, and managing the timeline. Google Nano Banana Pro generated the images: dark German streets, conspiracy corkboards with red string, an astronaut hiding a jar behind his back. And Google Veo 3.1 turned those stills into cinematic scenes with camera movement and zero-gravity shots. ElevenLabs handled the deadpan narrator.
It took one evening, no film crew, and no budget. A year ago the same project would have taken a small production company and a few weeks. The result came out rough around the edges, and I am no film producer. You can watch it here if you want to see what a midnight conspiracy-theory ad looks like. What I want to talk about is the tool at the center of that stack, the one that unlocked everything else: Veo 3.1.
What Veo 3.1 actually is
Released by Google DeepMind in January 2026 and expanded through April, Veo 3.1 is a text-to-video and image-to-video model integrated across Gemini, Flow, YouTube Shorts, and Vertex AI. Free at the consumer tier for any Google account. $19.99 a month on Google AI Pro gets you roughly 50 generations through Flow. On Vertex AI, developer pricing runs $0.40 to $0.75 per second.
What I saw working with it
Three things stood out from my night of production.
-
Creative iteration speed is the actual product. I generated around 30-40 clips to land the 15 that made the final cut. The prompt-review-reprompt cycle used to be a production day per shot. Now it takes the time it takes to type.
-
Google stack integration changes what the tool is. Veo is a capability woven into Gemini and Vertex AI, which means the model is built into tools many companies already run. My Nutella project was a Claude Cowork project orchestrating a Veo layer, then a Nano Banana layer, then an ElevenLabs layer. But for any business already running on Google Workspace, Veo 3.1 is already inside the building.
-
Cinematic language matters more than copy skill. Better results came from prompting like a cinematographer than a copywriter. Frame the shot. Name the lens. Specify the camera motion and the lighting. "Slow dolly push toward the astronaut. Anamorphic lens. Golden hour light breaking through warehouse windows." That kind of language pulls the model toward a look it can actually do well.
Where the jaggedness shows up
The same video model that nails a zero-gravity Nutella shot will sometimes give you an astronaut with seven fingers, or a warehouse light that moves in a physically impossible way between cuts. This is exactly the jagged intelligence pattern from this issue's terminology section, applied to video. Benchmarks tell you the average. Your specific shot tells you whether the tool is inside its frontier or outside it.
The practical response is not to wait for the jaggedness to smooth out. It is to match the stakes of the output to the jaggedness of the tool. Short-form content, internal training, pre-visualization, creative variants for A/B testing, those can absorb a retake. Final cuts of a corporate brand film cannot.
If you want to try it
If you have a Google account, you already have free access in the Gemini app. That is where I would start. For anything beyond small experiments, Google AI Pro at $19.99 a month is the sensible entry point. If you are thinking about Veo inside a marketing or training pipeline, Vertex AI is where the work should happen.
The honest use case for Veo 3.1 in April 2026 is short-form work. Marketing clips and social content, prototypes and pre-visualization, training videos, A/B creative variants, fake Nutella conspiracy ads built at midnight because a jar floated across Artemis II and you had an idea. Match the jaggedness of the tool to the stakes of the output. A 60-second ad can absorb a retake. A 90-second corporate film under final review cannot.
Your action step
Pick one short-form video idea you would never commission from a production team because the ROI doesn't justify the cost. A training clip, a social ad variant, an internal announcement, a product explainer. Try to build a rough 30-60 second version in Veo 3.1 this week. Prompt like a cinematographer, not a copywriter. Pay attention to how many iterations it takes to get something you'd actually use. That iteration count is the number you should use to price AI video into your next creative budget.
If you'd like to think through how generative video fits into your marketing or enablement stack, or want me to run a working session on creative AI for your team, I'd love to help.