Media Startup 'Every' Builds AI Agent to Clone Editor's Taste

Dan Shipper is the co-founder and CEO of Every, an AI publication that launched in 2020 as a bundle of business newsletters. [1] The company has roughly 30 employees and operates as both a publication and a product studio. [1] Every offers tools like an email assistant named Cora and a file organizer named Sparkle, which are bundled with its journalism for a $20-a-month subscription. [1] According to Shipper, AI now writes essentially all of the company's code. [1]

While humans still mostly write the company's essays, AI is changing the way that the company writes. [1] The company attempted to clone the taste of its editor in chief, Kate Lee. [1] Every collected a dataset of 30,000 of Lee's historical edits to build a copy-editing agent. [1] This new agent was then back-tested against her past work. [1]

Every's approach highlights a model where small media companies can leverage AI to scale both software development and specialized editorial taste. By automating code generation and building proprietary agents based on internal data, the company creates a unique hybrid of journalism and software product development that traditional publishers have rarely achieved. What the sources don't address: How accurately the copy-editing agent performed when back-tested against Kate Lee's actual historical edits.