Every, a 30-person media company that reviews artificial intelligence models, has trained an AI agent to replicate the editing style of its editor in chief, Kate Lee, using a dataset of 30,000 of her historical edits. The effort represents an unusual approach to scaling human expertise in the age of AI, capturing one employee's taste and judgment to distribute it more broadly throughout the enterprise.

What Is Every Trying to Accomplish With Its AI Clone?

Every, founded in 2020 by Dan Shipper and Nathan Baschez, started as a bundle of business newsletters and has evolved into a publication about artificial intelligence that draws significant attention across the tech industry. The company runs a roughly 30-person operation that combines journalism with product development, publishing columns like "Chain of Thought" and hosting a podcast called "AI & I," while simultaneously building software products including Cora (an email assistant), Sparkle (a file organizer), Spiral (a writing tool), and Monologue (a dictation app).

The copy-editing agent represents an attempt to solve a specific problem: how to maintain editorial quality and consistency as the company scales. Rather than hiring more editors, Every collected decades of Kate Lee's editorial decisions and used them to train an AI model that could apply her judgment to new work. The company then back-tested this agent against her past work to validate its effectiveness.

"We have tried to clone the taste of our editor in chief, Kate Lee, by collecting a dataset of 30,000 of her historical edits, using it to build a copy-editing agent, and back-testing it against her past work," Shipper explained.

Dan Shipper, Co-founder and CEO at Every

How Is Every Using AI to Increase Productivity?

• Code Generation: AI now writes essentially all of the company's code, allowing single engineers to run entire software products end-to-end, a capability that would not have been possible at scale before AI assistance became available.
• Editorial Distribution: By cloning the taste of experienced editors through AI agents, Every can apply consistent editorial judgment across more content without proportionally increasing its editing staff.
• Rapid Product Development: The company was able to launch and maintain six software products alongside a daily newsletter with minimal funding, a feat Shipper credits entirely to AI productivity gains.

Shipper noted that the company's growth trajectory would have been impossible without AI. When Every was 12 to 15 people, it was already running six software products and publishing a daily newsletter. "That's insane," Shipper said. "Even running a daily newsletter that grows, and that people like and read all the time, is hard. Then to add software products on top of that, without very much funding, it only became possible because we started to be able to get enough from a single engineer that you can have one person run an entire software product end to end".

The company has doubled in size over the past year, growing from roughly 15 people to around 30, even as it has automated more of its operations. This apparent contradiction reflects Shipper's view that AI is fundamentally limited by the data it learns from. "AI is trained on the residue of human expertise," he explained, "but can't see beyond it." In other words, AI can amplify and distribute what humans have already done, but it cannot generate entirely new directions or insights that don't exist in its training data.

Why Does Every Keep Hiring If AI Automates Everything?

The question of why companies continue to hire as they automate is central to understanding how AI will reshape work. Shipper's answer suggests that AI's limitations create ongoing demand for human judgment and creativity. Because AI learns from historical patterns, it cannot venture into truly novel territory. Every's journalists and product builders are still needed to identify new directions, test new models, and make editorial judgments that require taste and intuition.

Every's model also reflects a broader shift in how media companies might operate in the AI era. Shipper drew a comparison to The New York Times, noting that the Times was only able to launch specialized bundles like its games section, Cooking, and The Athletic after 150 years of operation and significant scale. "We can start to do that much earlier and more quickly, with less money," Shipper said, crediting AI for compressing the timeline and reducing the capital required.

The company's dual role as both a critic of AI models and a builder of AI-powered products creates an interesting dynamic. Every publishes critical reviews of frontier models, including a piece titled "A model pitched for everyone impresses no one" about Anthropic's Sonnet 5. Shipper emphasized that this critical stance is actually valued by the labs themselves. "They're asking, 'What do you think?' Because they want to make the model better, and they know that if we don't like it, it means something," he noted.

Shipper

"No one trusts a model company to tell you where they objectively sit. So we have a good position as an arbiter between them, and that's not something that model progress will get rid of," Shipper stated.

Dan Shipper, Co-founder and CEO at Every

Every's experiment with cloning editorial taste through AI suggests a future where companies capture the expertise of their most skilled employees and distribute it through AI agents. This approach could allow organizations to scale quality and consistency without proportionally increasing headcount, though it also raises questions about whether such systems can truly replicate human judgment or merely approximate it based on historical patterns. For now, Every's experience suggests that the answer lies somewhere in between: AI can amplify human expertise, but it still requires humans to set direction and make decisions that go beyond what the data shows.