The agent infrastructure wars just got real. What took weeks to build now takes minutes, and the companies solving deployment complexity are about to own the next phase of AI adoption.
01
Anthropic's Managed Agents eliminates the biggest pain point in AI development
Alex Albert from Anthropic says Managed Agents has become "both the fastest way to hack together a weekend agent project and the most robust way to ship one to millions of users." The service removes all the complexity of self-hosting AI agents while still allowing developers full control over tools, skills, and configurations. On VibeCode, developers can now go from prompt to deployed app without weeks of infrastructure setup.
Why it matters: Every startup currently burning engineering time on agent infrastructure just got leapfrogged by teams using Managed Agents. The competitive advantage shifted from "can you build it" to "how fast can you ship it."
Box CEO shows knowledge work automation in real time
Box CEO Aaron Levie demonstrated how their integration with Claude Managed Agents can automate document workflows in under two minutes. Using the Box API, teams can now set up automated document review, data extraction, and content integration with other business systems as background processes.
Why it matters: Your legal team reviewing contracts and your ops team processing vendor documents are about to get very productive. When document workflows run in the background, the bottleneck shifts from processing time to decision-making quality.
Google's Notebooks feature turns Gemini into NotebookLM
Google product manager Josh Woodward announced Notebooks, bringing NotebookLM's document analysis capabilities directly into the main Gemini app. Users can now upload 100 sources for free, organize chats into projects, and sync everything across devices. The rollout starts on web and moves from Gemini Ultra subscribers down to free users.
Why it matters: Google just made their best research tool available to millions more people. Every consultant, analyst, and knowledge worker who couldn't justify a separate NotebookLM workflow now has it built into their daily AI chat.
OpenAI's latest model tackles five mathematical proofs simultaneously
OpenAI CPO Kevin Weil shared that their newest models can solve multiple Erdős problems at once, with proofs that are "getting more elegant" as the models improve. Erdős problems are notoriously difficult mathematical challenges named after the prolific mathematician Paul Erdős.
Why it matters: Mathematical reasoning was supposed to be the hard ceiling for current AI. If models can handle multiple advanced proofs simultaneously, the gap between AI capabilities and human expertise in technical fields just shrunk dramatically.