Korva is a scroll of the internet’s best ideas.

A CEO is a glorified router
The company brain
Replit built an internal bot that became the brain of the company. Anyone can ask it about the business, the metrics, or where the company is headed. It doesn't look in one place: it pulls from Notion pages, a Postgres database, and the code base, and synthesizes everything into a report.
Hierarchy gives way
A self-driving data semantic layer connects and understands all of the company's different data sources, so the bot's answer reflects the whole picture. When information reaches everyone directly, hierarchy stops being the way it moves. The company takes the shape of a network, like an open-source project.
A self-driving company
The aim is to remove bureaucracy, so the running of the company happens automatically in the background and people can focus on the creative work. Replit calls the result a self-driving company. To invent the next thing, it also wants volume and velocity: many shots on goal.
How can a bot answer from many data sources?
Many company assistants use a technique called retrieval-augmented generation. The system first searches connected sources, such as documents and databases, for passages relevant to the question, then hands that text to a language model that writes the answer. The term comes from a 2020 paper by Patrick Lewis and colleagues at Facebook AI Research. Database questions are usually answered by turning the request into a query, such as SQL, and running it.
Why do organizations route information through managers?
A person can only keep up with a limited number of working relationships, so organizations group people under managers who pass information up and down. The sociologist Max Weber described this layered structure of offices as the core of bureaucracy in the early twentieth century. Fred Brooks observed in his 1975 book The Mythical Man-Month that the number of possible communication paths in a team grows much faster than its headcount. Hierarchy was one way to cut those paths down.
How do open-source projects coordinate without managers?
Most open-source projects coordinate in public: issue trackers and mailing lists let anyone read a discussion and reply to it. Eric Raymond's 1997 essay The Cathedral and the Bazaar contrasted this open style with software built by a closed team under central planning. The Linux kernel adds a layer of maintainers who review patches in their own areas before passing them upward. Contributors there are spread across many companies and countries.
More from Casey Newton
