Google held its annual developer conference on May 19th and, unlike a lot of these events, this one had some things worth paying attention to.
Gemini Spark: Google Finally Built a Personal AI Agent
The headline announcement for everyday users is Gemini Spark, a general-purpose AI agent, soon to be inside the Gemini app, that connects to your Google apps and reasons across all of them.
Spark can look at your Gmail, Calendar, Drive, and Docs all at once and actually do things with that information. You don’t have to copy-paste your emails into a chat window and explain who people are. It already knows.
Ask it to prep you for a Monday morning meeting and it can pull the relevant emails, find the shared Doc from last time, and give you a summary. That’s the vision, at least.
This is the most credible AI that actually knows your context product a major lab has shipped to a mainstream audience thus far. The bar for AI assistants has been “it’s only as good as what you feed it.” Spark is Google’s attempt to fix that problem by connecting the whole stack.
If you’re a heavy Google Workspace user, this is worth trying when it releases.
Gemini 3.5 Flash: The New Go-To for Builders
For developers, the more interesting announcement was Gemini 3.5 Flash.
Flash outperforms the previous flagship (Gemini 3.1 Pro) on nearly every benchmark, runs four times faster, and costs roughly one-third the price of comparable frontier models.
What makes this relevant beyond benchmarks is what it signals about Google’s strategy. They’re not just competing on capability anymore. They’re competing on the cost of running AI at scale, which is the actual bottleneck for most real-world applications. If you’re building anything production-facing with AI, Gemini 3.5 Flash just jumped to the top of the list to evaluate.
Gemini Omni: The World Model Worth Watching
Google also announced Gemini Omni, which they’re calling a world model — a model trained to “make things in the real world.” In their promotional material they say you can start with any video and use natural language to tell Omni to make whatever you want within the video.
It’s an interesting concept, but I’m skeptical about AI generated content as a default.
The Price Cuts Matter More Than People Realize
Google cut Google AI Ultra from $250/month down to $200 and introduced a new $99/month tier. Both tiers now include Gemini Spark, Gemini Omni, and the full model stack.
Pricing is an interesting topic, but one can probably assume that Google is trying to capture a wider share of the market of users who are willing to pay premium prices for the most capable AI tools.
Google Search Is AI First
Google also announced changes to the standard search results that we’ve known for a decade or more. Gone are the days of the list of blue links being the only result you receive back. For a year or so now they’ve been defaulting to the AI search results at the top of the page, but now when you hit search you’ll drop into an interactive AI chat.
Google says this won’t happen for every search, but it will happen to all users eventually. It seems like they’ll begin by prompting users to ask follow up questions prior to dropping into the full AI mode UI.
So What Do You Do With This?
If you’re a regular user: Plan to try Gemini Spark when they release it. The most useful thing you can do is connect it to your Google apps and give it one real task — something you actually spend time on, like prepping for a recurring meeting or triaging a cluttered inbox. See how much it can do without you having to explain the context manually. That’s the initial test. If it’s as powerful as they say, it could be your new personal assistant.
If you’re a developer or builder: Gemini 3.5 Flash API might be worth looking into for your projects rather than what you’re using now. Claude Code and Claude API seems to be a very popular model currently due to Opus 4.7, but 3.5 Flash might give it a run for it’s money.
One keynote doesn’t change everything. But this one certainly had some weight behind it.