Media company automates editing workflows with AI clone of its top editor

Every, a media and product studio led by CEO Dan Shipper, has built an AI agent trained on 30,000 copyedits by editor in chief Kate Lee to emulate her editorial taste and standards. Shipper described the effort as an attempt to capture a single employee’s expertise and distribute it more broadly across the company, framing it as a preview of how businesses may integrate AI with human judgment in the future.

The company also uses AI to automate software development and product operations, enabling a single engineer to ship and maintain entire products end-to-end. Shipper said this capability allowed Every to run six software products and a daily newsletter with a small team, and to double headcount from roughly 15 to 30 over the past year while publicly automating workflows.

Shipper argued that AI augments human expertise rather than replacing it, stating that AI is “trained on the residue of human expertise” and cannot see beyond it. He contrasted this with the company’s role as an independent arbiter of model quality, noting that no model company can credibly assess its own products and that Every’s critical reviews of frontier models help the labs improve while maintaining editorial independence.

The outlet’s “vibe checks” of new models—early access reviews of frontier systems—have at times created tension with labs like Anthropic, whose models Every reviews and on which it builds products. Shipper said labs often ask for feedback before publication and accept criticism as part of improving models, even when reviews are critical. He emphasized that Every’s goal is to help make better AI through partnership, not to “shit on” the labs.

Every bundles journalism with a suite of AI-assisted products—Cora, Sparkle, Spiral, and Monologue—under a $20-per-month subscription. Shipper acknowledged a “dirty secret” in writing with AI: many writers integrate AI into their workflows but do not publicly disclose it. He claimed that far more writers are using AI than admit it, while emphasizing that Every’s journalism remains primarily human-written.