When technology reporter Alex Heath has a scoop, he sits down at his computer and speaks into a microphone. He’s not talking to a human colleague—Heath went independent on Substack last year—he’s talking to Claude. Using the AI-powered voice-to-text service Wispr Flow, Heath transmits his ideas to an AI agent, then lets it write his first draft.
Heath sat down with me last week to showcase how he’s integrated Anthropic’s Claude Cowork into his journalistic process. The AI tool is connected to his Gmail, Google Calendar, Granola AI transcription service, and Notion notes. He’s also built a detailed skill—a custom set of instructions—to help Claude write in his style, including the “10 commandments” of writing like Alex Heath. The skill includes previous articles he’s written, instructions on how he likes his newsletters to be structured, and notes on his voice and writing style.
Claude Cowork then automates the drafting process that used to take place in Heath’s head. After the agent finishes its first draft, Heath goes back and forth with it for up to 30 minutes, suggesting revisions. It’s quite an involved process, and he still writes some parts of the story himself. But Heath says this workflow saves him hours every week, and he now spends 30 to 40 percent less time writing.
“I’ve always hated the zero-to-one process of writing a story … Now, it’s actually kind of fun,” he says. “Going out on my own, I realized I need AI to help with the volume.”
Heath is part of a growing contingent of tech reporters using AI to help write and edit their stories. The AI workflow is especially enticing for reporters who have gone independent, losing valuable resources like editors and fact-checkers that typically come with a traditional newsroom. Rather than just prompting ChatGPT to write stories, independent journalists say they are re-creating these resources with AI.
Their usage raises broader questions about the value of human journalists altogether. If people are using AI to write, edit, and fact-check their stories—what do humans bring to the table? A recent study from Google DeepMind researchers suggests that using AI in a lazy way can make your writing more homogeneous. It’s less creative, it has less voice, and it takes on a more neutral stance. To use AI well, journalists I spoke to say they need to understand why people are paying for their work in the first place. (WIRED’s policy prohibits the use of AI in writing or editing).
While some writers have built a career on their analysis and prose, Heath sees his value as his ability to get scoops. Claude makes it easier for him to spend more time chatting with sources and getting information out to his subscribers.
Several longtime journalists remarked to me that Heath’s workflow feels like a modern version of a long-standing institution: the rewrite desk. In the days before laptops and smartphones, reporters in the field would call in stories to a newsroom, where writers behind a desk would quickly weave those reported details into articles they could print for the next day’s paper. This allowed some reporters to spend their days covering events and talking to sources. In a way, Claude is now Heath’s rewrite desk.
“I feel like I’m cheating in a way that feels amazing,” says Heath. “I never did this because I liked being a writer. I like reporting, learning new things, having an edge, and telling people things that will make them feel smart six months from now.”
Jasmine Sun, who previously worked as a product manager at Substack, recently launched her own newsletter covering AI and Silicon Valley culture. Last week, she published an article in The Atlantic about how post-training makes AI models bad at writing by essentially beating out their creativity. Because of that, Sun never uses AI to write, but she has found promise using Claude as an editor.
Like Heath, Sun has fed Claude past articles she’s written and notes on her style. But she’s also instructed Claude to focus only on enhancing and developing her voice and taste, and never to be sycophantic. She tells Claude it “should never write a sentence for her. Your goal is to elicit out of Jasmine by providing feedback.”
Here’s part of the instructions Sun has shared with her Claude editor: “You are not a co-writer. You cannot perceive—you don’t have experiences, sources, scenes, or emotions to draw from. Your role is to help Jasmine write like the best version of herself—not just who she is on the page now, but who she’s trying to become as a writer. That means understanding both her current voice *and* her aspirations, including the writers and qualities she’s reaching toward.”
I asked Sun if she ever feels the urge to be lazy and just let Claude write for her. “I think [Claude] forces me to work harder than I would otherwise,” said Sun. “With a human editor, they're calling you on your bullshit. They're not letting you get away with lazy reporting or floppy prose.”
After speaking publicly about her use of Claude, Sun received criticism from people who were offended by the notion that AI could replace a human editor. Critics argued that AI can’t transform your ideas or challenge you as much as a human. Sun says she found the comments confusing. Most Substackers can’t afford to hire a human editor, so by adding Claude and instructing it to challenge her, Sun argues it’s made her process more rigorous. “To me, it is kind of like using Grammarly. You have a tool that says, ‘this sentence is bad,’ and then I have to go fix it,” she says. “[Claude is] more willing to tell me this entire section is bad and you should cut it. It's like a higher level of thinking and abstraction than Grammarly can do.”
Casey Newton, author of the newsletter Platformer, tells me AI has made him reassess the value of his publication. “I think there's an interesting distinction here. If the value is in the information, not the writing, then I think people will care less that AI did most of the writing,” says Newton. “If the value is in voice and opinion and argument and analysis, it seems like it's cheap to use AI to do the whole thing”
In recent years, Newton has focused on news analysis. But as AI improves, he says he’s changing his approach. "I actually need to shift the balance," he tells me. "I need to do less news analysis and more original reporting."
Newton is not using AI to write Platformer today but says he was inspired by Sun’s AI editor and has since tried to re-create it with a Claude agent based on his own articles. “I’ve just been really struck by it. At its best, its feedback is about as good as the feedback I've gotten from human editors,” says Newton.
Taylor Lorenz, author of the User Mag Substack, tells me she uses AI to help run her media business. She has Gemini come up with SEO-friendly descriptions for YouTube videos and has Claude help sift through data.
However, Lorenz says she’s not using AI to write or edit her articles today. She doesn’t trust AI systems with sensitive reporting materials and finds that AI just hasn’t been useful for writing and editing. She also just loves the craft of writing herself.
“I am a journalist because I like to help people understand the world and bring light to different issues,” says Lorenz. “I don’t want the AI to do that.”
Kevin Roose, a technology columnist with The New York Times, is using AI to help him produce a book about the race to build artificial intelligence. He claims AI tools have helped him shave two or three years off the process.
Most recently, Roose tells me, he created a team of Claude agents to help edit his book, led by a “Master Editor” agent. Other sub-agents are in charge of things like fact-checking, making sure the book matches his writing style, and offering positive and negative feedback. (To be clear, he’s still working with human editors, too.)
However, Roose has not handed over the writing of his book to AI just yet. Like Sun, Newton, and Lorenz, Roose feels that he’s still just better at writing than an AI model. “I think the models tend to be fairly generic and depersonalized, but also, I like doing this,” he tells me.
Roose makes clear that he’s anything but a skeptic—he expects AI models to get better than him at everything eventually. But he is human, and for now that’s an edge. “I am not under some romantic illusion that I possess a special, irreplaceable perspective. But what I am is a person, and I think that for now people, at least some people, like hearing from people,” he says.
This is an edition of Maxwell Zeff’s Model Behavior newsletter. Read previous newsletters here.
