Notes on Android development, AI, agents, new things I'm learning, and whatever else is worth writing about. Mostly for myself, but you're welcome here.
Every major AI model writes the same way: one word locked in before the next begins. Google's DiffusionGemma throws that rule out and writes an entire block at once instead. Here is how it actually works, in plain English, and what it costs you in return.
How A2A — the open standard already backed by AWS, Microsoft, Salesforce, and 150 more organizations — lets AI agents from any company, any framework, discover and work together over plain HTTP.
Google announced Managed Agents at Google I/O 2026. It lets you deploy a fully functional AI agent with a single API call, no servers to manage, no infrastructure to build. Here is what it is, how it works, and why it matters even if you have never written a line of code.
At Google I/O 2026, Google released ADK 2.0 with a graph-based execution engine and collaborative multi-agent workflows. Here is what changed, what it means, and why it matters even if you have never written an agent before.
Gemini Embedding 2 is Google's first AI model that converts text, images, audio, and PDFs into the same number space. I built MemoryVault to understand how it works and why it changes how we think about search.
Most AI demos show one agent answering one question. This is different. Three specialized agents fire simultaneously, a fourth synthesizes everything into one ranked answer, and an Android app calls the whole system live. Google ADK, Vertex AI, Cloud Run, Docker. The full picture.
How I used Gemma 4 E2B via LiteRT-LM in my Recipe Scanner app to translate and structure recipes in 140+ languages, entirely on the phone, and what I learned along the way.
My personal system for staying on top of the AI explosion as an Android developer, without spending all day on Twitter.