Agency Wins: Gemini Omni and the GPT-5.5 Autonomous Lab

Google just unveiled Gemini Omni at I/O 2026. Last year, Nano Banana brought us image generation. Today, Omni gives us video-first native multimodality, taking audio, images, and text to generate physics-aware video on the fly.

A new video model is impressive. The real systemic shift, however, happened in a cloud laboratory in Boston.

OpenAI connected GPT-5.5 to Ginkgo Bioworks and built an autonomous science loop. Over 150 hours, the agent ran 36,000 physical reactions to optimize cell-free protein synthesis. It brought the specific production cost down by 40 percent without a single human in the loop.

This is the line in the sand. We are moving beyond chat interfaces and into raw agentic deployment. Intelligence is no longer measured by how well a model answers a question. It is measured by how well it persists in a physical environment, designs its own experiments, and executes them while the human operators are asleep.

Cynicism is easy. Building is hard. Build anyway.

The human bottleneck is completely gone. We are now the architects, not the executors. This is the floor. It is the worst the technology will ever be, and it advances every hour.

Steal it, remix it, share it. Tell me what you are building, or tag @UsAndAI if you have already started.