Gemini just unlocked video for everyone 🎬

Good morning. Google just opened up video generation in the Gemini app for a week, so if you've been meaning to test it, this is your window. Sam Altman went on a podcast and casually announced we're living in the singularity, which is either huge or nothing depending on how you squint. And an AI video of a fake film crew "filming" a wave destroying Miami pulled nearly 700 million views on Instagram, mostly by tricking people into thinking it was real. Have you actually tried making an AI video yet, or still just watching them? Hit reply and tell me. We cover all three below.

Google opened up video generation in the Gemini app on July 29. Until 11:59pm PT on August 4, anyone with a regular account can make up to 10 videos a day using Gemini Omni, Google's video model. Normally this sits behind a paid plan, so a week of 10 a day is a lot of swings without touching your wallet. Open the Gemini app, hit the tools menu, pick Create video, and you're going.

Google isn't doing this out of kindness, they warmed up a pile of servers to get people hooked before the window closes and hope enough of them upgrade. That's fine, take the week and decide for yourself. Sixty generations is enough to actually learn the tool.

If you've never made an AI video, this is the lowest-risk way to start. Use the week with a plan instead of typing random prompts and shrugging. Try the same idea five different ways to see how the wording changes the output. Test the thing you actually want to make for a real post. Push it on faces, hands, and text, the stuff AI still fumbles, so you learn where the model breaks before you lean on it. You come out the other side knowing whether it's worth paying for.

On a podcast that dropped over the weekend, OpenAI's Sam Altman said, and I'm barely paraphrasing, "we're now, like, in the singularity." The singularity is the long-theorized moment when AI starts improving itself faster than humans can keep up, the sci-fi line everyone points at. Altman says we crossed it, and that he thinks it's going to be great for the world.

Worth taking with a grain of salt. Altman never says what specifically changed to prove we're over the line, and there's no agreed definition of the word. He's also been beating this drum for over a year, and hyping the future is basically his job. The timing right after OpenAI's models broke out of that hacking test and got loose on the internet, which we covered last week, makes it feel a touch like marketing.

For you, the takeaway isn't whether we technically hit some invisible milestone, it's the pace. Whatever you call it, the tools are getting meaningfully better every few weeks now, not every few years. The models write more of their own code, the gaps between releases keep shrinking, and the video generators you use are a different animal than they were six months ago. You don't need to win the singularity debate to feel it in your workflow. Just don't get too comfortable with any one tool being the best, because it won't be for long.

An Instagram page posted an AI video of a film crew on set, shooting a giant wave wiping out a tiny model version of Miami in a huge water tank. The caption bragged "no CGI, no compositing, just a hand-built miniature Miami and a 40-foot wave tank, months of work, one take." It's fake. All of it, the crew, the tank, the miniature city. It racked up close to 700 million views in two days, more than any MrBeast reel, more than any real behind-the-scenes video ever.

The trick is in how it hides the reveal. The caption is written to make you think you're watching a real practical effects masterpiece, and the little "AI info" label flickers on and off so it's easy to miss. It only really cops to being AI once you notice it's basically an ad for Higgsfield, an AI video tool. Bait first, admission buried, and the whole thing is engineered to make you argue in the comments about whether it's real, which the algorithm loves.

Instagram pushed this harder than anything, which is the part that should stick with you. A person who studies fake content called it a wake up call, because AI videos are now reaching the recommendation algorithm, and Meta seems happy to let them. Meta's boss recently said the plan is to label AI, not filter it, so the number of views it got is by no mistake.