Keep pulling the thread on Logan Kilpatrick.
Gemini Omni is a single, unified model capable of handling text, audio, music, and video, not a system that routes requests to multiple specialized models like Lyria or Veo.
Approximately 350,000 Android apps were built in Google's AI Studio in the first week after the feature launched.
Google's "Agentic Gemini era" is powered by the Anti-Gravity Agent Harness, which serves as a connecting through-line for all Google products.
The Antigravity agent harness powers agent features in other Google products, including Search, the Gemini app, Google Cloud, and AI Studio.
Contrary to initial assumptions, the introduction of AI-powered answers has been "incredibly positive sum" for Google Search, leading to increased search activity.
Google acquired Windsurf to build the Antigravity platform, realizing it needed a dedicated product for long-running agentic coding to improve its coding models.
The coding performance improvements in Google's Gemini 1.5 Flash model were achieved entirely through post-training techniques, not a new pre-training run.
A team at Google used agentic coding to deliver the Gemini macOS app faster than any team had ever delivered a Mac app at the company.
In October 2023, Logan Kilpatrick predicted on Twitter that everyone would be able to "vibe code" video games by the end of 2025.
The first version of Google's Gemini Omni model to be released is the "Omni Flash" model, with its best-performing capability currently being video editing.
Google DeepMind collaborates with partners across the Google ecosystem, including Android, Google Cloud, Gmail, and Workspace, to deploy Gemini to billion-user products.
Prior to the introduction of Gemini, there was no common AI through-line connecting Google's approximately 50 products.