Search is not broken but economically constrained. Google’s product works well, but its $400B search business prevents it from offering cheap, unbundled APIs, which creates an opening for smaller players despite inferior scale. This is true because offering flexible search access would cannibalize its core revenue.

Agent-driven search changes the product requirements entirely. Humans want fast, simple results, while agents generate complex queries with filters and tolerate latency for higher-quality outputs. This leads to fundamentally different architectures where latency, ranking, and verification trade-offs shift.

The AI stack is becoming modular and interchangeable. Models, search providers, and tools can be swapped depending on performance and use case, which forces builders to design systems that can evolve quickly. This is true because capabilities change within months, making static architectures obsolete.

AI collapses execution time across functions. Tasks that previously required teams and weeks can now be executed by a single person in minutes using coding agents, enabling one-person workflows across engineering, sales, and operations. The constraint shifts from execution to judgment and problem framing.

Geography still matters because of talent density and ecosystem behavior. Zurich works due to concentrated AI talent from Google and ETH, but San Francisco accelerates faster because startups, capital, and early adopters cluster tightly. This creates a compounding advantage where distribution and experimentation happen faster.