Korva is a scroll of the internet’s best ideas.
Everyone has an internal Twitter feed
Notebook channels
Anyone can join anyone else's, including those of researchers, and read whatever they want. Posts range from updates to provocative ideas. Leaders use theirs to scale their beliefs as the company grows quickly: a post on the importance of being comfortable leaving money on the table was read by every engineer who has since joined the growth team, and it differed from what many of them had known before.
Open debate
The channels are open to argument too. After an all-hands, an employee who disagreed with something Dario had said posted in his notebook channel, and it sparked a big debate. Challenging leadership publicly is encouraged, which builds trust.
Feeding Claude
Some of this feeds Claude as well. Certain onboarding documents carry a note from the HR team to check with a specific person before editing, because it is a document that Claude references.
Why does public disagreement build trust?
When disagreement is punished, problems stay hidden until they are large. Harvard professor Amy Edmondson studied hospital patient-care teams in a 1999 paper in Administrative Science Quarterly. She found that the teams with a more open climate reported more errors, because their members felt safe talking about mistakes. Leaders who invite challenge in public show everyone else that speaking up carries no penalty.
Why do companies write culture down at scale?
Anthropologist Robin Dunbar's research suggests people can maintain roughly 150 stable relationships. Beyond that size, beliefs can no longer spread by word of mouth alone, because most colleagues never meet. Written channels let a leader's view reach thousands of people in the same words. Many fast-growing companies publish internal handbooks and memos for this reason.
Why would a document be flagged for Claude?
When an AI assistant retrieves company documents to answer questions, their wording becomes part of what the model reads, so an edit can change its answers. Retrieval-augmented generation was described in a 2020 paper by Patrick Lewis and colleagues at Facebook AI Research. A note asking editors to check with a named person keeps that document from changing without its owner knowing.
More from Lenny's Podcast: Product | Growth | Career
