Whenever tech correspondent Alex Heath uncovers an exclusive, he settles in at his workstation and addresses a microphone. He isn’t conversing with a human coworker—Heath became a freelance writer on Substack last year—rather, he’s engaging with Claude. Leveraging the AI-driven voice-to-text platform Wispr Flow, Heath relays his concepts to an AI assistant, subsequently allowing it to compose his initial manuscript.
Heath met with me recently to demonstrate how he has incorporated Anthropic’s Claude Cowork into his reporting methodology. This artificial intelligence utility is linked to his Gmail, Google Calendar, Granola AI transcription platform, and Notion records. Furthermore, he has developed a sophisticated skill—a personalized collection of directives—to aid Claude in composing text in his unique manner, encompassing the “ten guiding principles” of emulating Alex Heath’s prose. This skill incorporates prior pieces he has authored, guidelines on his preferred newsletter layout, and observations on his tone and literary approach.
Claude Cowork subsequently automates the composition stage that previously unfolded within Heath’s mind. Once the assistant completes its preliminary draft, Heath engages in an iterative exchange with it for as long as half an hour, proposing modifications. This constitutes a rather intricate procedure, and he continues to pen certain segments of the narrative personally. Nevertheless, Heath asserts that this operational method conserves him numerous hours weekly, and he now dedicates 30 to 40 percent less effort to writing.
“I’ve consistently disliked the initial genesis phase of constructing a story … Presently, it’s actually quite enjoyable,” he remarks. “Venturing out independently, I recognized the necessity of AI to manage the sheer quantity of work.”
Heath belongs to an expanding group of technology journalists employing AI to assist in drafting and refining their articles. The AI-driven process is particularly appealing for correspondents who have become freelance, thereby forfeiting valuable assets such as editors and fact-verifiers commonly found in a conventional news organization. Instead of merely prompting ChatGPT to generate narratives, autonomous journalists state they are replicating these functionalities using AI.
Their application provokes wider inquiries regarding the overall worth of human journalists. If individuals are utilizing AI to compose, revise, and verify their narratives—what unique contributions do humans offer? A recent investigation by Google DeepMind scholars indicates that employing AI indolently can render one’s prose more uniform. It exhibits reduced creativity, possesses less distinctiveness, and adopts a more impartial posture. To effectively leverage AI, journalists I conversed with emphasized the need to comprehend the fundamental reasons why individuals remunerate their efforts. (WIRED’s guideline forbids the utilization of AI in authorship or revision).
While certain authors have forged careers based on their analytical insights and literary style, Heath perceives his primary contribution as his knack for securing exclusives. Claude facilitates his ability to devote increased time to conversing with informants and disseminating data to his readership.
Numerous veteran journalists commented to me that Heath’s operational method resembles a contemporary iteration of an enduring establishment: the rewrite department. During the era predating laptops and mobile phones, field reporters would relay stories telephonically to a news agency, where desk-bound writers would swiftly interlace those reported specifics into pieces suitable for the following day’s publication. This arrangement enabled certain journalists to dedicate their days to reporting on occurrences and engaging with informants. In essence, Claude now functions as Heath’s rewriting station.
“I sense I’m gaining an unfair advantage in a manner that feels extraordinary,” states Heath. “I never pursued this profession because I enjoyed being an author. I appreciate investigative journalism, acquiring novel insights, maintaining a competitive advantage, and informing individuals of facts that will make them perceive themselves as astute half a year hence.”
Jasmine Sun, who formerly served as a product director at Substack, recently initiated her personal newsletter, delving into AI and Silicon Valley’s societal norms. The previous week, she released a piece in The Atlantic discussing how post-training intrinsically diminishes AI models’ writing proficiency by effectively stifling their inventiveness. Consequently, Sun abstains from employing AI for drafting purposes, yet she has discovered potential in utilizing Claude as a proofreader.
{content}
Source: {feed_title}