Korva is a scroll of the internet’s best ideas.
Build for where AI is heading
Getting named by chatbots
Search engine optimization for AI is already starting, and the open question is how large it can get. If the ambition is making sure a client's name appears in answers from ChatGPT, Meta AI, Gemini, or Claude, that is a pretty small ambition and a pretty small outcome.
What agents change
Agents change the size of the prize because they take actions. When an agent decides whether to buy from Amazon, Target, or Walmart, that decision sits in a very big part of the conversion funnel that now needs to be optimized, and it is more than presenting a link after a navigational search.
The ambition of the team and the quality to expand as AI capabilities expand decide which market it is. Startups aimed at where AI is today tend to get steamrolled by the labs themselves, while those aimed at where it is heading, with agents controlling computers and making full purchase decisions for consumers, face a huge addressable market.
Why do thin AI startups get overtaken?
Developers have a name for this pattern: being Sherlocked. Apple's Sherlock search tool added features in 2002 that matched a third-party app called Watson, and Watson's maker, Karelia Software, became the standard example. When a platform owner folds a popular add-on into the platform itself, the add-on's reason to exist disappears. AI labs can do the same when a new model release turns a standalone feature into a built-in capability.
How do AI agents operate a computer?
A computer-use agent takes screenshots of a screen, reads them with a vision model, and then issues mouse and keyboard actions such as clicking and typing. It repeats that loop until the task is done. Anthropic released this capability in public beta in October 2024, and OpenAI released its Operator agent in January 2025.
How did search optimization begin?
Search engine optimization began in the mid-1990s, when website owners tuned pages for early engines such as AltaVista. Google launched in 1998 with PageRank, a method developed by Larry Page and Sergey Brin at Stanford that ranked pages partly by the links pointing to them. Link-building and page tuning grew into an industry around those ranking signals.
More from The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
