Google I/O keynote: Stratechery

The first takeaway from both keynotes is just how impressive Google’s infrastructure is.

which the quality of an AI-mediated experience is dependent on the AI being integrated into the problem space and its relevant data.

Gmail, can help you summarise the email thread and you can also ask questions which it will search through all of your emails and give you an answer to. When typing an email response, it has contextual suggestions which you can simply send.

What makes the demo particularly compelling is the fact it happens in the Gmail mobile app: Gemini’s summarization and compiling capabilities are even more useful when you consider that the smaller screen size of a smartphone would make the manual alternative even more painful.

Gemini works at the system level, so instead of going to a separate app, I can bring Gemini right to what I’m doing. Now we’re making Gemini context aware so it can anticipate what you’re trying to do and provide more helpful suggestions in the moment.

In the demo Burke brings up Gemini as an overlay to generate an image, summarize a YouTube link, and interrogate a PDF attachment; in every case he is able to do so without, as he noted in the opening, needing to invoke a separate app, import information, etc. It’s a good demonstration of how being “context aware” can provide a much better experience than having to context switch, and it’s only possible because Google controls the operating system within which all of this information is being accessed. Once again these are features that will roll out in the next few months.

hat part is really important as we prototype these experiences. We’re thinking hard about how to do it in a way that’s private, secure, and works for everyone. These are simple use cases, but they give you a good sense of the types of problems we want to solve, by building intelligent systems that think ahead, reason, and plan all on your behalf. The power of Gemini with multi-modality, long context, and agents brings us closer to our ultimate goal, making AI helpful for everyone.

Second, the way in which multi-step reasoning results are displayed are quite conducive to advertising — which, by the way, can be that much more targeted the more fleshed out a query is. In other words, to the extent Google can train its users to leverage multi-step reasoning the better it could turn out to be for their business.

’s a similar story with the planning capability: it should be pretty straightforward to insert ads in a way that makes sense, and there are obvious opportunities for building out a fully-contained shopping experience, with all of the targeting and attribution possibilities entailed in that.

Finally, what was striking to me about the “inspiration” SERP was the extent to which it feels more like a feed than a results page. Again, this makes sense: if you’re looking for inspiration you’re in more of an exploratory mode than a research one, and this new format could be more compelling.

Here Are 52 of My Favorite Quotes