The website that created an AI clone of its editor in chief
Platformer Casey Newton
Every built an AI copy-editing clone of its editor in chief. That’s a neat trick until you realize the company sells both the critic and the thing being criticized.
Based on reporting by Platformer, Casey Newton — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Every has spent years doing something a little unnerving for the AI age: reviewing the tools it also depends on, then shipping products built on the same models. Dan Shipper, the company’s co-founder and CEO, says that tension is part of the business. Every publishes criticism, runs “vibe checks” on new frontier models, and still needs those same labs to keep improving the stack underneath its products.
The company started in 2020 as a bundle of business newsletters. It has since turned into a publication focused on AI, with Shipper’s Chain of Thought column, the AI & I podcast, and a steady stream of early model reviews. But Every is also a product studio. Its offerings include Cora, Sparkle, Spiral, and Monologue, and they’re all packaged with the journalism in a $20-a-month subscription. Shipper says AI now writes essentially all of the company’s code, while humans still mostly write the essays.
The most striking example of that shift is inside the editing workflow. Shipper said Every tried to clone the judgment of editor in chief Kate Lee by collecting 30,000 of her historical edits, building a copy-editing agent from them, and testing it against her past work. The point isn’t to replace her taste; it’s to spread it. If Lee can’t spend all day copy editing launch emails, landing pages, and articles, the model can at least approximate her standards while she moves on to other work.
That logic runs through Shipper’s explanation of the company’s growth. Every roughly doubled from about 15 people to around 30 over the past year, even while trying to automate everything it can. He argues that AI is trained on the residue of human expertise, which means it can handle a lot of familiar problems but still needs experts when the work gets specific, messy, or new. In other words: the better the tools get, the more the company needs people who know where the tools fall short.
And that’s where Every seems to think its real leverage lives. Model labs can build the ovens, Shipper said, but that doesn’t mean they know how to make the soufflé. Every wants to be the place that tells people which model is actually good, then uses those models to build things the labs may not get around to themselves. It’s an awkward business model. It’s also, for now, a workable one.
My take — AI-written commentary, not fact-checked reporting
This is the new AI power move: not “replace the editor,” but bottle the editor and ship the bottle. Companies love to call that augmentation because it sounds tasteful; it’s also a very clean way to turn judgment into software and keep the credit loop running. The uncomfortable truth is that the best AI businesses may be the ones that know exactly how to commoditize their own experts without saying the quiet part out loud.
Read more about this at: Platformer