Google’s New AI Update Just Shocked The AI Industry – Gemini 3.5 Pro , Gemini Omni, Gemini Spark

"Ségou just released"
ok strong work, guess google just won, by calling it self segou xD

Terrible update and they stung us AI pro users. Flow is dire. Credits gone. Also bugs and huge wait limits! Not happy

This was a really interesting watch. I’m actually pretty excited about how hard Google is pushing into agents with stuff like Gemini Spark and Antigravity 2.0. The idea of something running in the background and actually getting things done without me having to constantly check on it feels like a real step forward. 

The speed of 3.5 Flash stood out too. At the same time though, a lot of these demos still feel pretty polished and controlled, so I’m curious how well any of this will actually hold up once people start using it for messy, real-life work. Still, it does feel like we’re slowly moving toward agents that could actually be useful instead of just looking cool in a presentation.

Thanks for the updates, but what I just saw related to Antigravity 2 is a cocky company that has no clue on how engineers work. They stripped the VSCode based model, added background agents which even Hermes has now, reduced all features and functions to a Manus type screen, give you a crap flash model which is useless for real world work, so that they can hit a ridiculous price point of 1k USD api tokens when engineers use pro models to save $$, stripped, nerfed the Antigravity 1 which engineers were using with a pro model that was crippled, then call it 2.0; Codex is far more robust, gives you substantial tokens, though less features than Cursor, has the agents and most importantly gives significant use of thinking level at reasonable token use; Cursor is improving daily and has all features and functions for an engineer; It would be nice if you weren't always "Oh this is great!" when in actuality it is just Google moating and nerfing their product

Editing tools are cool but the output quality is far behind competition. I will not use omi for video production quality China models are beating them right now on that. But omi tools can be used.

Essa atualização Google em todas as minhas janelas agora não dá nem para conversar conversa um pouquinho e já dá tranco

Just wanted to say thank you for these AI updates and all your hard work. I mainly use GPT, but Gemini is my go-to for troubleshooting. After watching your video, I tried the updated Gemini and noticed a real improvement in interaction. I'm currently building a multimodal interactive AI audiobook that's 98% accurate with an original sci-fi novel, handling all questions, image, and video generations. This new Gemini update was so stable, I didn't even need to make any changes, modifications, or calibrations, so they did an amazing job on this model.

Google Chrome now has a built-in AI called Gemini Nano. Four gigabytes. It runs directly on your computer and sees everything you do in the browser. It can read your files, remember your data, and act on your behalf: browse websites, fill out forms, download files. All of this is legitimate, signed software that is not flagged as a virus because it is not a virus. It is a standard feature.

The danger is straightforward. Whoever controls the Chrome update server can send a command to your browser at any time. The browser will silently download illegal content from the internet — child pornography, for example, or classified documents, or terrorist communications generated by another AI, it does not matter what. It will place those files in your Downloads folder, tweak the metadata, and create logs that mimic your activity. That is it. There is now criminal evidence on your computer.

After that, an anonymous tip is triggered, or the provider's automated monitoring system detects "suspicious traffic." The police arrive, seize the computer, and find the files. You say, "I was framed." The forensic report shows the files were downloaded from your IP address, during your working hours, under your account. There are no signs of a breach because there was no breach. You go to prison. Everything is legal.

There is no weak link in this mechanism. The agent is already in the browser. Remote control through automatic updates is already working. The legal framework for secret requests from authorities to corporations already exists. The tool for planting evidence — generative AI — is already mature. The local executors — the police and the courts — are already functioning.

You cannot stop using the browser because your entire life depends on it. You cannot prove the files were planted because all the digital evidence will point to you. And you cannot blame hackers because no hacking occurred. Everything was done by legitimate, signed, trusted software.

This is a tool of absolute, unprovable, remote power. It does not require tanks, camps, or show trials. It simply deletes an inconvenient person the way you delete an unwanted file. Quietly. Without noise. Without consequences for the one who gave the order.

Omni will be integrated into remixing YouTube shorts…

Gemini 3.5 flash is soooo expensive 😳

(Apple intelligence) Siri can now create you an agenda by you just saying 'hey Siri create me an agenda " 😮

As someone who had been researching protein folding in the past, the details in the video got me irritated a bit