The real 'next big thing' would be integrating an engine like Gemini with OS-level hooks (similar to the OpenClaw approach) so agents can manipulate app windows and state directly. Resurrecting Web Intents as 2-way App Intents would be the key to making this work.
Also, keeping prompts as local .md files with an Obsidian-like system editor would be a huge win for power users. Simply gating Gemini behind 'premium' Chromebooks feels like the old 'licking the cake' strategy from the Google+ days—trying to force a new product's success by coopting existing hardware rather than building a superior platform.
I can imagine having Gemini + local Gemma working with Agents, which have access to my e-mail (ideally on GMAIL, but also supporting outlook), keeping local history of my visited sites and messages... and using RAG or something even better, ideally with looking also on repos I have checkouted to my file system, and maybe even whole file system....
Work related e-mail about "sending invoice to customer"... it may suggest proper content for e-mail. Having "dashboard" with summary of todays communication to you, your tickets (at work) and so on....
Can Google build such thing? If somebody can - it will be them. Will they build it? Probably not, they would prefer to build 3rd version of Google Pay.
