Google launches Gemini 3.5 Flash and shifts focus to AI agents

In short: Google launched Gemini 3.5 Flash, a new AI model meant to power “agents” that can do multi-step work, not only answer questions.

What happened

Google announced Gemini 3.5 Flash at its I/O 2026 developer conference. Google says it is its strongest model so far for writing code and for running “AI agents,” which are AI helpers that can plan and carry out tasks, like a digital assistant that can actually do the work and not just suggest what to do.

According to Google, Gemini 3.5 Flash can run complex coding workflows on its own. In internal tests, it could even build an operating system from scratch, which is the core software that makes a computer run.

Google also introduced Antigravity 2.0, a desktop app built for this agent-style way of working. Google says Flash is designed to be fast because agent tasks can involve many steps running at once, like several coworkers working on different parts of the same project and then combining their work.

Google said Gemini 3.5 Flash is now the default model in the Gemini app and in AI Mode in Google Search globally. The company also said it plans to bring more agent features to Search, including tools that let people create and manage agents.

Why it matters

For regular people, this could change what “using AI” feels like. Instead of typing a question and getting an answer, you may ask AI to handle an ongoing job, like tracking something you care about and updating you, or helping complete a longer task over hours. It also raises safety questions, since AI that can act on its own can cause more harm if it makes a bad choice, and Google said it has added stronger safeguards.

Source: TechCrunch AI