Agentic AI. That is the game now. Google just launched Gemini 3.5 Flash at I/O on Tuesday. No waiting. It is live for everyone right this second.
Sundar Pichai called it a game changer. Inside the company anyway. He claims it runs at half the cost of what the competitors charge. Speed is the currency here.
Next month, Gemini 3.5 Pro drops. They are testing it now. The promise is parity with OpenAI’s GPT-5 and Anthropic’s Claude. But with a twist. It acts. Not just talks.
“It can handle sessions lasting multiple hours.”
That is the difference. Most models get tired after a paragraph. Gemini Flash wants to run all day. Long-term projects. Whole research arcs.
The Always-On Helper
Meet Spark. Think of it as an OpenClaw challenger. Or a rival. OpenClaw blew up earlier this year by letting AI live on your hardware. Mac Minis selling out like concert tickets.
Spark? It lives in the cloud.
Josh Woodward from Google Labs is rolling this out slowly. First, trusted users. Then, a beta for the folks who pay extra. Ultra subscribers get priority. Always the story with AI.
It acts on your behalf. Real directions. Not just “write a poem” but “check my inbox and organize the meeting.”
Testers are using it to track school schedules. Plan parties. Watch for urgent emails while you sleep. Woodward described the feel like tossing something over your shoulder and having a safety net catch it.
Chrome will connect it to the rest of the internet by summer. Third-party apps included. It will go everywhere you browse.
Coding Gets Serious
Google has played catch-up on code. Anthropic has Claude Code. OpenAI has Codex. Those two set the standard for developer tools.
Now, they are changing Antigravity. It was just a platform before. Now, it is agent-first.
Koray Kavukcuoglu calls it a place to manage teams of AI workers. Not one bot. A squad. One agent builds the website. Another designs the logo. A third plans the product rollout.
There is a desktop app now. And a command-line version. It is trying to become the headquarters for your digital workforce.
The App Gets Pretty
The interface gets a name change too. Neural Expressive. Sounds expensive. It involves new animations, haptic taps, and better colors.
Gemini will show images when it can. Timelines. Visuals. Less text dumping. More visual answers.
They are putting all this weight into one place. Flash models. Omni models. All integrated.
It is a bet that we don’t just want chat bots anymore. We want hands. Digital ones, sure. But hands nonetheless.
What happens when your AI stops asking what to do?
