I've been working on stuff outside of Google, which is nice because it lets me build things purely because I want to. No sprint planning, no product requirements, just me and a problem I want to solve.
The big one is GeminiFit Coach! I lift weights and cycle, and every fitness app I've tried either nails one thing or is mid at everything. So I'm building my own. It's an Android app that hooks into Health Connect, Strava, and Gemini AI. The idea is one dashboard for all my fitness data: weight, nutrition, lifting progress, cycling stats. Gemini gives me actual recommendations based on MY data instead of generic "drink more water" advice.
I wrote up a full design doc for it in September 2025, and it's the most fun I've had on a project in a while. No constraints! I'm the user, I'm the PM, I'm the engineer. If a feature doesn't make sense I just don't build it.
I've also been sketching out something called "Project Phoenix," basically a productivity app that would pull together Google Tasks, Calendar, Keep, and Docs into one view with a GitHub-style contribution heatmap. I keep calling it "agent-first" because I wrote the spec as a set of instructions that an AI coding agent could execute. Which is either very 2025 or very unhinged. Possibly both.
Building for yourself hits different. At work I'm shipping to millions of users and that's its own kind of satisfying, but scratching your own itch? Building the exact thing you wish existed? That just feels right.