Model Releases
Reporting from Google I/O 2026 with the four biggest themes from one of the biggest AI labs in the world. š¤ Voice AI as an interface Googleā¦
Reporting from Google I/O 2026 with the four biggest themes from one of the biggest AI labs in the world. š¤ Voice AI as an interface Google and Samsung announced new Gemini-powered glasses with Gentle
Reporting from Google I/O 2026 with the four biggest themes from one of the biggest AI labs in the world. š¤ Voice AI as an interface Google and Samsung announced new Gemini-powered glasses with Gentle Monster and Warby Parker (congrats @NeilBlumenthal!). You can tap the side of the frame or say "Hey Google" to summon Gemini for real-time translation (I tested Korean), navigation, photos, and contextual search about whatever you're looking at. They also released Docs Live, which lets you verbally brain-dump a full document and edits (even with fillers, tangents, mid-thought changes) and Gemini writes it for you. More āliveā products coming from Google soon. The era of typing back and forth in one chat thread has been dead for over a year, my friends. š¤ Agent-first everything Gemini Spark is a 24/7 agentic assistant that runs on dedicated VMs in Google Cloud, so you can close your laptop (this got a big laugh) and it still functions. It has its own Gmail address you can email tasks to, pulls files from Drive, and will plug into other tools like Uber, OpenTable, and Zillow. Rolls out to AI Ultra subscribers next week. This is Googleās answer to Open Claw. Youāll also soon seen agentic coding inside Google Search itself. ā»ļø Orchestration efficiency Google released Gemini 3.5 Flash and Google CEO, Sundar Pichai, said it runs ~4x faster than other frontier models at less than half the cost of comparable frontier alternatives, allowing businesses worried about token spend (which is, news flash, most enterprises) to spend more wisely. š World models Demis Hassabis announced Gemini Omni, the beginning of Googleās real push into world models. Omni lets you not just generate videos, but also edit them in natural language. āHey Omni, make my outfit head to toe leopard printā āGreat taste, Allie.ā Demis said eventually Omni will allow you to input anything and output anything. Big statement there tbh, with no timeline. ----------- I was also lucky enough to personally meet with the CTO of @GoogleDeepMind and the Chief AI Architect at Google, @koraykv, where we discussed AGI, data centers, world models, and more. Video goes live on YouTube next week: https://www.youtube.com/@AKMofficial Notably absent from the event were two topics Iām tracking: (1) self-learning and (2) collaboration. Not a single demo showed two people sitting in the same AI workflow or system building something together! And no mention of recursive at all, which feels like the big theme in Silicon Valley. Personally, I worry AI is becoming single-player mode only. I want it to be easier for AI to bring people together.
Source: Allie K. Miller (X) | 2026-05-20