What Google just dropped three of ⚡

Good morning. Google put out three new models this week and one of them is off limits to normal people. Adobe's camera app will now look at your shot and tell you what you did wrong, which is either helpful or brutal depending on the day. And Disney's new kids cartoon is blowing up mostly because the internet decided it was AI slop, except that's not really what happened. Has anyone accused your stuff of being AI yet? Hit reply and tell me. We cover all three below.

Google dropped three new Gemini models on July 21. The main one is Gemini 3.6 Flash, their new everyday workhorse. The second is Flash-Lite, a stripped down version built to be fast and dirt cheap. The third is called Flash Cyber, that one hunts for security holes in code, and Google is only giving it to governments and a few trusted partners. Also yes, only the first one got the 3.6 name. The other two are still 3.5. Google naming things has never made sense.

These aren't really smarter than the last batch, they're just less wasteful. Same job, less thinking out loud, so they cost less to run. The main one got a price cut and Flash-Lite is cheaper by a lot, fast enough that answers come back basically instantly. Google also admitted the upgrade everyone was waiting on, Gemini 3.5 Pro, isn't ready because it missed their own quality bar. Gemini 4 is already in the oven though.

For creators, these models can watch video and look at images, not just read text. So the cheap one is good enough for the boring half of your job now. Captions for a stack of clips. Tagging your b-roll so you can actually find things later. Scanning a long video to pull the best hooks out of it. Titles and descriptions for a week of uploads. You're pointing it at the hundred little chores you're doing by hand right now.

Adobe added a set of AI features to Project Indigo, its experimental iOS camera app, on July 20. The headline feature is photo critique: shoot something and the app gives you a professional read on your framing, lighting, color, and emotional impact. It also does capture and edit suggestions, telling you how to reshoot the thing, and then how to fix the version you already have using specific Lightroom controls. The project is led by the guy who built the Pixel camera, which explains a lot.

Most AI editing tools make you guess at a prompt until something works. This one is just buttons, so you press a thing and get the same result every time instead of rewording "make it cinematic" for ten minutes. It'll also wipe objects out of your shot, fake that blurry background look, and drop different styles on top. Funny detail: it's running on Google's Nano Banana model instead of Adobe's own Firefly.

The reason this matters beyond photography is the coaching loop. Every AI tool you use right now executes and hands you a result. Almost none of them tell you why the result is weak or what you'd do differently next time, which is exactly the skill gap that separates people making AI slop from people making things that hold up. One catch: it's a limited free test on a small slice of users in version 1.1, edits cap at 2K, and it's only sticking around for a few weeks.

Disney Jr quietly put out a show called Ozzy Fox on YouTube and YouTube Kids with no announcement at all. Two episodes, a family of cartoon foxes, songs about doing chores and using the potty. It pulled around 800,000 views in two days. It also got accused of being AI slop almost immediately, with people pointing at props that look unfinished, mouths that don't match the words, and the fact that there's no readable writing anywhere in the backgrounds. Those are the usual signs of AI video. That's not what this is.

Disney made it with a French studio whose whole thing is AI assisted animation. Real artists built the characters and drew the storyboards and the important poses by hand. The AI does two narrow jobs, both trained only on this show's own artwork: it turns a rough sketch into a posed character, and it draws the filler frames between the poses an animator already made. That second one has been automated in animation for years. It took the show from months of work down to weeks.

The accusation stuck anyway, and it stuck hard, because "AI slop" has stopped meaning "a robot made this" and started meaning "this looks a bit cheap to me." If you're using AI anywhere in your process, and after last week's YouTube memo you're probably already thinking about it, hiding your tools won't save you. Clean lip sync, readable text in frame, props with all their parts. People are hunting for those specific tells, and every one of them is fixable. Nobody disclosed anything here either, and the audience least able to ask questions about it is four year olds.