Gemma 4 12B comes to laptops

Google wants multimodal AI on your laptop

Gemma 4 12B can take audio and images without separate encoders, which is the real news here.

Google says the new 12B model runs locally on laptops with 16GB of VRAM or unified memory. It also supports native audio input, uses an Apache 2.0 license, and is available through tools like LM Studio, Ollama, Hugging Face, Kaggle, llama.cpp, and vLLM.

This matters because Google is making local AI feel less like a demo and more like a developer workflow. The encoder-free design is the part I’d watch, since lower latency and lower memory use are what decide whether people actually use these models at 11 p.m. with a laptop fan running.

The broader move is clear: smaller models are being asked to handle more types of input without leaning on cloud infrastructure. That could make privacy-sensitive audio and vision agents easier to build.

The question is whether developers trust the local performance enough to ship real products.

Netflix wants less scrolling

Netflix thinks generative AI can make choosing a show feel less like sorting through a crowded shelf.

Elizabeth Stone, Netflix’s chief product and technology officer, said at Bloomberg Tech in San Francisco that AI could make the viewing experience “more personalized, more interactive, more immersive.” The goal is to help customers deal with content overload, which is still one of streaming’s most ordinary problems.

This is practical. Netflix does not need AI to feel futuristic here; it needs AI to make the home screen less tiring at the end of the day, when the TV light is on and nobody wants to keep searching.

The risk is that better recommendations become more steering than helping. Netflix already shapes what people see, and generative AI could make that influence harder to notice.

The signal is that streaming competition is moving deeper into the interface, not only the catalog. The next fight may be over who makes choosing feel easiest.

Google wants AI to suggest your life

Dreambeans reads across your Google apps while you sleep, then hands you AI-illustrated story suggestions in the morning.

Google Labs launched the iOS and Android app for eligible U.S. Google AI Ultra subscribers. With permission, it pulls from Gmail, Calendar, Photos, YouTube, and Search History to produce a daily batch of about 10 to 14 stories.

The coffee angle is real. Google says the app is meant to feel like a small morning serving of ideas, with recommendations for places to visit, topics to explore, events, trips, or articles.

I think the product is less strange than the name, because Google has been moving toward personal AI that acts on your existing data. The harder part is trust, since even good suggestions can feel invasive when they come from your inbox, photos, and calendar.

This could become useful for people who already live inside Google’s apps every day. It could also remind users how much Google already knows.

Would a useful suggestion make that feel better, or worse?

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.