Recently, I had the privilege of attending Google’s I/O developer conference with colleagues Will Rinehart and Shane Tews. The event featured (literally) 100 AI announcements and live demonstrations, including Waymo and Wing drone delivery. Reflecting on the two days, several themes emerged reshaping technology and society.
AI Integration Everywhere
Analyst Ben Thompson said he was overwhelmed by the scope of everything announced, yet underwhelmed by the lack of standalone products. But I think this is exactly what Google was demonstrating—innovation is not just about standalone products, but how deeply AI will be integrated into daily tools. Rather than separate platforms, Google’s vision integrates AI directly into the services you already use: search, office tools, video conferencing, and even wearable technology like smart glasses.
The last decade was defined by the Internet of Things (IoT), where devices (like appliances, vehicles, and sensors) were connected via the Internet. We’re now entering the “Intelligence of Things” era, embedding advanced AI directly into everyday systems and devices. Navigating this shift will demand not only technical expertise, but extensive collaboration across various sectors (and with policymakers), a point repeatedly emphasized by Demis Hassabis, CEO of Google’s DeepMind.
Proactive AI: Anticipating Your Needs
Another recurring theme was Google’s vision for AI, which is “personal, proactive, and powerful.” Personalization is familiar—technology shaped around your preferences. But the real leap is proactivity—AI systems that anticipate your needs and take initiative on your behalf. This shift is crucial as it enhances productivity, reduces friction in daily tasks, and fosters a more seamless and intuitive interaction between technology and users.
Unlocking Deeper Insights with Integrated Research Tools
For many leaders and policymakers, their initial encounter with AI agents will be through Deep Research assistants. Gemini 2.5 Deep Research breaks down your research project into tasks, explores the web, and synthesizes findings. You can now turn your Deep Research report into a web page, podcast, or interactive quiz.
Immersive Communication
One of the most remarkable technologies I demoed was Beam. It’s impossible to describe—even the demo video doesn’t do it justice. I knew I was looking at a screen, but I felt like I was talking with a real person. That realism is achieved not just from the 3D experience, but from subtle cues like eye contact and facial expressions, elements typically lost on platforms like Zoom.
Google Meet will also provide live translation capabilities, enabling real-time multilingual conversations. It functions much like a translation service; the original audio is dialed down and you hear the translation in your language but, most impressively, it is in your own voice.
Revolutionizing Learning Through AI
Google emphasized AI’s potential to dramatically enhance education and learning. Liz Reid, Google’s head of Search, noted a trend: Search queries are becoming increasingly more complex because AI allows users to ask more sophisticated and layered questions.
One of the most compelling demos was Project Astra, a multimodal AI assistant. Google demonstrated its capabilities by facilitating a bike repair—a task that blends diagnosis, problem-solving, and hands-on execution but above all—learning during each step. Astra seamlessly combined visual analysis, speech, contextual search, tutorial retrieval, and even custom how-to video generation using Veo 3. It offered a powerful glimpse into the future of AI-enabled learning and tutors: interactive, timely, multimodal, and deeply intuitive.
Google also released additional research on LearnLM, which used a “learning arena” with educators and pedagogy experts who conducted over 1,300 blind comparisons of five leading AI models across learning scenarios. Gemini 2.5 Pro was preferred in over 73 percent of expert evaluations in aligning with core pedagogical principles, including managing cognitive load, fostering metacognition, and adapting to student needs. Additional assessments confirmed Gemini’s strength in key tutoring tasks, including mistake identification, short-answer grading, and grade-level text adaptation.
Notably, the LearnLM model is now integrated into Gemini 2.5 and other products, embedding these advanced educational capabilities directly into the broader AI model rather than existing solely as a specialized standalone education LLM.
A Human-Centric Future: Technology Serving People
A central theme was the vital role of agile public policy in enabling the safe testing and deployment of advanced technologies like self-driving cars, drones, and AI. Excessive regulatory hurdles risk stalling and stifling innovation. Thoughtful, forward-looking policies can unlock these technologies’ potential to improve safety, productivity, and quality of life.
Perhaps the most important reflection came during the opening performance by Toro y Moi, whose AI-powered live music set helped set the tone for the event. He concluded his set by saying, “We’re here today to see each other in person, and it’s great to remember that people matter.” The delicate balance between cutting-edge innovation and the human element was a reminder that AI’s power lies in its capacity to serve and uplift people.
Below is an audio summary of this blog post: