The AI infrastructure landscape is splitting in two: companies building tools for humans versus tools for AI agents. And the early winners are betting everything on the latter.
01
Vercel CEO: "We used to build tools for humans, now we're building them for agents"
Vercel CEO Guillermo Rauch announced his company is expanding Vercel Labs with a mission to build developer tools specifically for AI agents. The team has already shipped agent-focused tools like agent-browser, portless, and json-render that have racked up 22.8 million downloads. Rauch frames this as a fundamental shift in how developer tools are designed.
Why it matters: If agents become the primary users of development tools, every piece of software infrastructure needs to be rebuilt. Vercel is betting that the companies who redesign for agents first will capture the entire next generation of AI-powered development.
Every founder Dan Shipper discovers the killer feature hiding in plain sight
Dan Shipper shared his experience using analytics platform PostHog directly inside his AI coding tool Codex. Instead of switching between apps, Codex can write database queries, analyze the results, and even kick off agents to write pull requests based on what it finds. His conclusion: "a browser inside your desktop coding tool beats an agent in your browser."
Why it matters: This is the workflow revolution nobody saw coming. When your AI can seamlessly jump between code, data analysis, and production changes without context switching, the entire concept of separate software tools starts to break down.
Together AI partners with Adaption to streamline model fine-tuning
The companies integrated Together's fine-tuning capabilities directly into Adaption's data optimization platform. Teams can now optimize datasets, run fine-tuning, evaluate results, and deploy custom models without switching between multiple services.
Why it matters: Fine-tuning is moving from a specialist skill to a standard workflow. When data scientists can customize AI models as easily as they run A/B tests, every company becomes an AI company.
Developer Zara Zhang predicts the end of user interfaces
Zhang shared her shift from building web apps with traditional interfaces to just sharing GitHub repositories that people customize themselves. She connected this to Andrej Karpathy's concept of "idea files" where you write down concepts and let people's AI agents build customized implementations.
Why it matters: If AI agents can build interfaces on demand, why spend months designing them for humans? This could be the beginning of the end for traditional UX design.
Product manager Peter Yang spots the next personal AI breakthrough
Yang praised a technique for personal agents from Josh Pigford, calling it "brilliant" and saying he's adopting it for his own AI project OpenClaw. The specific implementation wasn't detailed in the post.