AI Agents, Enterprise Shifts & Claude Code: May 2026 Builder Insights

One of the most thought-provoking observations this week came from Swyx, who : "the agentic excel thing is basically what u get when you expand the side panel to be the main thing." This deceptively simple observation captures a profound shift in how software is being reimagined.

Vercel CEO Guillermo Rauch added his own signal to this trend with a cryptic but telling about "The ideal 🪩"—suggesting the perfect interface is one that feels almost invisible. This aligns with Rauch's long-standing philosophy that the best developer experiences remove friction, and agents represent the ultimate friction-removal layer.

Product managers and designers need to start thinking about "agent-first" interfaces rather than "AI-enhanced" ones. The distinction matters. An AI-enhanced interface adds intelligence to existing workflows. An agent-first interface redesigns the workflow entirely around natural language intent and autonomous execution.

Aaron Levie, CEO of Box, delivered what might be the most important thread of the week. In a , Levie described a conversation with a Fortune 500 pharma CEO who "commented on how much more technical talent they need right now." The key insight? The demand isn't shrinking—it's transforming.

Levie articulated a critical mismatch: "Right now there's a temporary mismatch between the jobs that used to be sought after in some fields and the new jobs that becoming in demand in those fields." For years, the CS pipeline was built to produce engineers who would "build customer-facing software in some form." Now, organizations across every industry need technical talent to implement and manage agentic systems.

This isn't just a Silicon Valley phenomenon. As Levie noted, he's "hearing from every CIO and CEO across nearly every industry right now" that technical talent is desperately needed—but the nature of that technical work has fundamentally changed.

Peter Yang, who runs a popular AI newsletter with 140K+ readers, at a "big week for all my Google friends," suggesting enterprise AI products are about to get major updates. This aligns with the broader pattern: every major platform is racing to embed agentic capabilities into their enterprise offerings.

"One of the best things students and colleges can do is not bail on learning and teaching the fundamentals of any given domain. AI will trick you into thinking you don't need to go deep in a particular area, but that's wrong."

Levie's argument is nuanced. The expert with AI will always outperform the novice with AI. The experienced software developer who's "built and scaled complex systems using agents" will outrun someone who's "just vibe coding." The designer who understands visual hierarchy and user psychology will build better products with AI than someone prompting blindly.

Levie perfectly: "Those that can steer AI agents properly, figure out how to evaluate their work, fix their mistakes, and incorporate their work into a workflow will always be the most potent users of these tools."

Levie also acknowledged that "despite some of the rhetoric in the valley that this is less [important] now, that couldn't be further from the case." This is a healthy correction to the hype cycle. Yes, AI is transformative. No, it doesn't mean you can skip learning the hard stuff. The most valuable professionals in 2030 will be those who combined deep domain expertise with fluency in AI agent orchestration.

YC President Garry Tan fired off the week's most energetic take in a that directly challenged cost-cutting orthodoxy: "Ken Griffin doesn't understand the ceiling just got raised. Some 20-something maybe reading this will build the cracked AI-human-computer-symbiosis team that will supersede his whole operation because he is too distracted about lowering cost."

Tan's message to founders? "Boil the ocean don't cut your costs." This is a deliberately provocative framing. While traditional business wisdom emphasizes capital efficiency and lean operations, Tan is arguing that we're in a rare moment where the upside of bold experimentation vastly outweighs the downside of overspending.

The reference to a "20-something" building a team that could supersede a multi-billion dollar operation isn't hyperbole—it's Tan reading the same signals every top builder is reading. When tools like Claude Code can serve as a second brain and AI agents can handle entire workflows, the leverage available to small, talented teams is unprecedented.

The most technically fascinating content this week came from the AI & I podcast, which featured Noah Breyer's remarkable Claude Code setup. As the host described it, Breyer "rigged a home server in his basement, put his Obsidian vault in it, and then runs Claude code on top"—enabling him to "think, research, write, and even ship code right from his phone."

This isn't just a cool hack. It represents a fundamental reimagining of what AI coding tools can be. Claude Code, in Breyer's workflow, functions as "a true second brain, a thinking partner that asks him sharp questions, pulls research from his whole note archive and the web, and even keeps a running log of what he's learned and what his best ideas are."

Thariq, who works on Claude Code at Anthropic, a lighter take on the multi-device workflow many developers are adopting—"dual wielding" laptops to maximize their AI-augmented productivity. This humorous observation (which earned ) hints at a serious trend: developers are integrating AI tools into every facet of their workflow, not just code generation.

The "second brain" pattern will likely become standard for knowledge workers within two years. Imagine every consultant, researcher, and strategist with their own AI-augmented knowledge system that remembers everything they've ever read, thought, or written—and can synthesize insights on demand.