Why Anthropic Is Hiring a $320K Copywriter

Opening

Dear reader, an intriguing job posting went up last week. Anthropic, the AI company, is looking for a copywriter — with a salary of up to $320,000. That’s about ₩440,000,000 (~$320,000) in Korean won.

But Andrej Karpathy¹, who recently joined that very same company, published a tool just two months earlier analyzing the AI-replacement risk of 342 U.S. occupations. In it, copywriters scored 8 to 9 out of 10 for exposure — placing them among “the job categories most vulnerable to AI.”

So the same company is now offering a record-breaking salary for a job it identified as ripe for AI replacement. Is this a contradiction? To cut to the chase: the answer to this paradox is that “writing” and “figuring out what to say” are entirely different skills.

🔍 Two Signals From the Same Company

In March of this year, Karpathy released an interactive tool that scored the AI exposure of 342 occupations on a 0-to-10 scale, based on U.S. Bureau of Labor Statistics (BLS) data. The core criterion was simple: “Can this job be done from home with just a laptop?” If so, the score climbed.

Copywriters, writers, and editors scored 8 to 9 — the same range as software developers (8 to 9) and management analysts (9). Roofers and hairdressers, by contrast, scored 0 to 1. The takeaway: the more a job involves producing digital output in front of a screen, the higher its exposure to AI.

Yet on May 22, that same Anthropic posted an opening for a Copy Lead to handle enterprise-customer messaging strategy — salary range $255,000 to $320,000. A separate Head of Copy and Content role was also open, paying $320,000 to $400,000. Converted to Korean won, that tops out at ₩550,000,000.

The qualifications section of the posting has some striking language: “Someone who, given a brief, comes back with an idea rather than a draft.” “Someone who sets direction through creative judgment and taste, but who’s also willing to do the hands-on work.” And: “Someone who has designed a brand voice as a system.”

What stands out here is that Anthropic isn’t asking for someone who’s “good at writing text.” It’s asking for editorial judgment — the ability to make complex technology clear, concrete, and readable.

🧭 Why Frontier AI Companies Are Investing in Their Voice

This isn’t just an Anthropic story.

In April of this year, OpenAI acquired the tech talk show TBPN². Launched in late 2024 and airing live every day since early 2025, the show had only 58,000 YouTube subscribers — yet it was a channel where Big Tech CEOs like Mark Zuckerberg and Satya Nadella made appearances. Its ad revenue was roughly $5 million in 2025 and was projected to top $30 million in 2026. OpenAI acquired the company for hundreds of millions of dollars and folded it into its strategy team.

A company that builds AI models is buying a media company, and offering ₩400 million to a copywriter — a job it classified as ripe for AI replacement. Why is this happening?

As I see it, the answer lies in a shift in the bottleneck. Frontier AI companies’ technical capabilities are already plenty strong. The performance gap between ChatGPT, Claude, and Gemini is narrowing to decimal points on benchmarks. At this point, the axis of competition is shifting — from “building a better model” to “conveying to the market why this model matters.”

Anthropic’s creative studio landing page has this line:

“We care about craft³, making things by hand and using AI when it makes sense.”

It’s telling that an AI company describes its own writing philosophy using the phrase “making things by hand.” This isn’t mere rhetoric — it means these companies themselves know best that text generated by AI and messages designed by humans are not the same thing.

📐 ‘Writing’ and ‘Figuring Out What to Say’ Are Different Skills

Karpathy’s AI-exposure assessment isn’t wrong. AI is very good at the act of generating text itself. Drafting emails, first-pass press releases, social media copy — AI already handles a large share of this “writing work.” When a copywriter said back in 2023 that “my job is already being replaced by ChatGPT,” that wasn’t an exaggeration.

But that’s not what Anthropic is paying ₩400 million for. If you’ve used AI services yourself, you know that just churning something out, or stamping out more of the same, is something AI can do faster and in unlimited quantity.

Reread the job posting and the core competency it’s after becomes clear: “A campaign headline is different from a white paper’s introduction, and a white paper’s introduction is different from paid ad copy. You need to be able to calibrate each of these while maintaining a single voice that runs through all of them.” That’s not “text generation” — it’s “message architecture.”

To sum up the distinction:

• What AI is good at: rapidly generating text that fits given conditions. Mimicking tone, following structure, hitting a word count.
• What AI isn’t good at: deciding what story to tell. Designing the order, the metaphors, and the emotional texture through which complex technology should be conveyed to non-experts. And keeping that design consistent across the very different surfaces of a campaign, a product page, an event, and social media.

Put simply, the cost of writing is converging toward zero, while the price of the judgment needed to figure out what to say is actually rising.

What’s interesting is the BLS’s own forecast. It projects that writing-related jobs will grow 4% from 2024 to 2034 — roughly in line with the average growth rate across all occupations (3.1%). That number cuts against the intuition that writing jobs should shrink in the AI era. What that 4% signals is that the writing that’s disappearing and the writing that’s newly emerging are fundamentally different in nature.

What’s disappearing is clear: formulaic writing with a fixed pattern — templated press releases, first-draft product manuals, SEO blog posts. AI already does this plenty well. What’s becoming more expensive is equally clear: writing that translates complex technology into the language of the market, designs a brand’s voice, and varies a single message across multiple channels. Anthropic’s ₩400-million job posting is the price tag on the latter.

Zooming out, this trend isn’t unique to writing. According to an analysis by The VC Corner, between mid-2025 and early 2026, six CTOs from billion-dollar companies — including Workday, Instagram, and Box — joined Anthropic not as executives but as individual contributors (ICs). They gave up their leadership titles to go back to being researchers. When the most expensive talent in the AI industry is all moving in the same direction, that direction is pointing at something real.

Anthropic President Daniela Amodei majored in English literature at UC Santa Cruz. In an ABC News interview this past February, she said: “As AI does more and more of the work, the things that make us human will matter more.” She added: “What we look for when we hire at Anthropic is people who are excellent communicators, highly empathetic, and want to help other people.”

An AI company led by a literature major is paying the highest price for writing. That’s part of why I write this newsletter — and it’s no coincidence.

Oz’s Lens

Honestly, this phenomenon doesn’t feel foreign to me at all.

In my work on technology strategy, I’ve watched this pattern repeat itself. As technology matures, the bottleneck to market adoption shifts from the product to the message. Early on, whoever has “the better technology” wins. But once the technology gap narrows, the contest shifts to “how well you can explain that technology to the market.”

Cloud in the 2000s and SaaS in the 2010s both followed this pattern, and the same shift is happening in the AI market right now. Anthropic’s ₩400-million job posting and OpenAI’s media acquisition are the price tags on that shift.

One thing not to misread here: saying “writing matters in the AI era” doesn’t mean “you need to write good prompts.” What Anthropic is looking for is someone who can build the skeleton of a story and keep a single message consistent across multiple channels. This isn’t about tool proficiency — it’s about instinct and judgment for language.

As the cost of generation approaches zero, the value of deciding what to generate goes up. Anthropic’s job posting is the price tag on that value.

Closing

To sum up: what AI replaces is the act of writing; what it can’t replace is the ability to figure out what to say. Judging what story to tell, in what order, and with what sensibility, is still the domain of humans. And no one knows this distinction better than the companies building AI.

Is there a part of your own work that you feel AI could never replace? Let’s talk about it in the comments.

References & Further Reading

Primary sources

• Anthropic, “Copy Lead, Enterprise” job posting, 2026. : This is the job posting that sparked today’s newsletter. Its qualifications section reveals what Anthropic actually means by writing ability.
• Anthropic, “Copy & Content Lead, Launches” job posting, 2026. : Detailed requirements for the Head of Copy and Content role (salary $320,000–$400,000).
• Andrej Karpathy, “US Job Market Visualizer”, karpathy.ai/jobs, March 2026. : An interactive tool scoring the AI exposure of 342 U.S. occupations on a 0-to-10 scale. The live page has since been taken down, but the open-source code remains on GitHub.
• CNBC, “OpenAI acquires popular tech podcast TBPN”, April 2, 2026. : An article covering the background and deal size of OpenAI’s TBPN acquisition.

Background

• Business Insider, “Anthropic cofounder says she doesn’t regret her literature major”, February 15, 2026. : An article on Daniela Amodei’s humanities background and her views on human capabilities in the AI era.
• U.S. Bureau of Labor Statistics, “Writers and Authors: Occupational Outlook Handbook”, 2024-2034 outlook. : The BLS’s official forecast that writing occupations will grow 4% even in the AI era.
• The VC Corner, “Breaking: Andrej Karpathy Joins Anthropic”, May 19, 2026. : Covers Karpathy’s move to Anthropic alongside an analysis of the talent-absorption pattern in which six CTOs of billion-dollar companies joined Anthropic as individual contributors.

The author, Kwangseob Ahn, is a professor of business administration at Sejong University and lead consultant at OBF (Oswarld Boutique Consulting Firm). He teaches statistics and data analysis — business data management and business analytics — while leading GTM and AI strategy consulting in the field, designing the seam between technology and business. He has published academic research on a memory architecture for AI dialogue systems (HEMA) and runs Daily Arxiv, a daily curation of global AI papers. He holds a master’s from Korea University’s Graduate School of Technology Management and a KMBA. He is the author of Homo Brainless: The People Who Outsource Their Thinking.

Footnotes

1. Andrej Karpathy: A co-founder of OpenAI and Tesla’s former head of AI. He joined Anthropic’s pretraining team in May 2026. He’s also known for coining the term “vibe coding.” ↩

2. TBPN (Technology Business Programming Network): A daily live tech talk show centered on Silicon Valley, launched in late 2024. Hosted by John Coogan and Jordi Hays, it was acquired by OpenAI in April 2026. ↩

3. Craft: Here, this means something closer to craftsmanship than mere “skill” — a meticulous work ethic that hand-tunes every detail. ↩