AI is learning to build biology

AI is learning to build biology

Scientists fed two AI models nearly nine trillion letters of genetic code and asked them to invent a biological virus from scratch. The result: sixteen functional bacteriophages, none of which exist anywhere in nature. Researchers used AI models called Evo 1 and Evo 2, built and tested about 300 designs in the lab, and found that some AI designs sometimes worked better than nature’s own. That’s a very specific kind of “move fast” energy.

Gobble's Take: When AI starts out-performing nature in a lab, the polite assumption that it’s “just autocomplete” looks a little thin. Source: Perplexity Search

AI is moving from model demos to operating discipline

This week’s AI pulse says the industry is moving beyond model adoption toward the operational realities of running AI at scale. Google is turning Maps into an agentic transaction layer, Airbnb is using AI to compress product-development cycles, Meta is deploying multi-agent coding across large repositories, and LangChain is building infrastructure for persistent agents. On top of that, Canva and Databricks are exposing the economics challenge behind widespread AI usage, while Anthropic’s move into custom chips and its massive compute commitment reinforce that intelligence now depends on infrastructure.

Gobble's Take: The shiny part was never the hard part; the hard part is now billable, governable, and impossible to ignore. Source: Perplexity Search

Robots, local inference, and creators all get pulled closer to the edge

This week in AI also points to a shift toward real-world data and consumer hardware. Dyna Robotics is using massive volumes of human video to teach robots physical skills, with the research suggesting robots may learn more efficiently without relying solely on robot-specific training data. Meanwhile, MiniMax H3 is coming to Mac through an open-source inference engine, and YouTube is changing its Partner Program for 2027 with new monetization opportunities across Premium Lite, Shorts, shopping, and brand partnerships.

Gobble's Take: The center of gravity is drifting outward: from cloud-only AI to robots, Macs, and creator wallets. Source: Perplexity Search

In Case You Missed It

Yesterday's top stories:

• OpenAI's GPT-5.6 family goes broad — after a very controlled detour
• NVIDIA, Apple, and Google all showed up with shiny new toys
• Meta's Muse Spark 1.1 arrives with aggressive pricing and a new business model