Get Your Hands Dirty

Today we’re testing a new newsletter format, aimed at giving our readers both a taste of our long-form writing and our perspective on what matters in AI today. Let us know what you think.—Kate Lee

Today’s top story

“Your Best AI Strategy Starts at the Top” by Natalia Quintero and Mike Taylor: Executives might be waiting on the sidelines to see what will happen with AI, but they need to be getting their hands dirty with the tools, write Natalia Quintero and Mike Taylor, both part of Every’s consulting team. That’s because AI can’t be evaluated like software, where you compare features, platforms, and integrations. It needs to be treated like a new kind of employee.

Natalia and Mike offer five concrete things for executives to do this quarter—starting with suspending skepticism—to get started building AI-native organizations. Read more.

Signal

Anthropic’s OpenClaw ban is a gift to OpenAI

The news: Anthropic blocked Claude subscriptions from being used with third-party agent harnesses like OpenClaw. OpenAI hasn’t.

The context: Anthropic’s stated reason for the ban is to prioritize compute for its own products, saying flat-rate subscriptions weren’t built for the high usage of third-party tools.

It’s a valid argument: Agents that run 24/7 are enormously expensive. But rival OpenAI has raised so much money it can afford to let subscribers use their models however they want.

The implications: Anthropic’s ban provides an opening for OpenAI to siphon away users. The strategy appears to be working: Opus 4.6 token usage is significantly down week over week; GPT-5.4’s has surged.

Bigger picture, the future of the industry depends on figuring out ways to drive down compute costs. (Running frontier AI agents like OpenClaw can cost $300–$1,000 a day, a number that’s only growing.)

OpenAI has a clear advantage here. It’s building its own data centers, which puts it closer to the metal on compute. Meanwhile Anthropic is buying compute from third parties, and will never have as low a cost basis.—Laura Entis

Good prompts don’t guarantee good design

Your prompt might be excellent, but most AI image generators follow instructions without aesthetic judgment. Recraft V4 is different. It combines exceptional prompt accuracy with visual taste. Create photorealistic mockups, native vector logos, or editorial layouts—production-ready on first pass. Crisp text in any language. Composition that holds up. Design teams at Ogilvy, HubSpot, and Airbus already use it.

Try Recraft V4

Want to sponsor Every? Click here.

New job alert

We’re flagging new job postings that signal where AI is reshaping teams.

Anum Hussain at Ashby, a recruiting technology company, is hiring a “Lead, Content Library.” The idea is to treat the company’s existing content like a product: Organize it, resurface it, track what’s losing viewership, and make sure the right piece reaches the right person at the right moment.

• What’s been true: Content teams hired people to make more new content.
• What’s changed: AI makes production cheap, so the new challenge is to get maximum value from content that already exists.
• What’s new: This role only makes sense when one person can manage a much larger body of work—and with AI, they can.—Katie Parrott

Inside Every

AI adoption has a before and after—the aha moment is the line

People talk about “technical” and “non-technical” when it comes to AI adoption, but that distinction is getting less useful by the day. The more revealing split is between people who have had the AI aha moment and people who haven’t. Once you’ve crossed that line, the question isn’t whether you’re technical enough for AI—it’s what you want to build with it.

That’s why getting to that aha moment is such a key step—and that magic moment is different for everyone. Our consulting team says that a typical aha moment for clients in using Claude is getting a daily digest of the overwhelming stream of communication—Slack, email, Jira, etc. On a recent episode of our podcast, Kate Lee, Every’s editor in chief, says her aha moment was when was feeling overwhelmed by managing the hiring process for several key roles earlier this year. Though she did look at every application, AI helped do a first pass on the hundreds she received, and offered “a way to evaluate everyone against consistent criteria.” She also used AI to set up all the job descriptions in Notion.—Eleanor Warnock

Who’s the author when AI does the writing?

In book publishing, the “author” and the “writer” aren’t always the same person. The author is whose ideas drive the work (generally, the name on the front cover). The writer is whoever puts them on the page (sometimes credited, often not). A celebrity might be the author of their tell-all memoir, but their ghostwriter is the writer.

AI has made everyone else confront this distinction. If someone uses AI to write a book, can they call themselves the author? When we spoke about this recently at Every, my colleague Mike Taylor‘s instinct was no—to him, authorship requires suffering. The pain of thinking something through is inseparable from the work itself. That framework applies in some contexts. But publishing already has a working answer: The person with the ideas is the author, full stop.

The harder question is: Which part of authorship do we care about—having the idea, doing the writing, or suffering enough for both? Mike’s frame isn’t entirely wrong, but perhaps slightly mislocated. As a former literary agent, my view is that the suffering doesn’t disappear when AI does the drafting (just ask Katie Parrott); it’s just even more likely to show up in the self-judgment—the nausea you feel when something you’ve published isn’t as good as you wanted it to be.—KL

Steal this workflow

Workflows we’ve tested and liked—ready to drop into your own process.

If you’re designing something new, Claude Code can generate working pages, full design systems, and clickable prototypes in minutes. Where it falls short is the last mile—the small decisions that make something feel made. Every designer Benjamin Osemwengie puts it this way: HTML gets you to good. A canvas-based tool like Figma gets you to great.

Try it this week: Generate the system, structure, and first-pass pages with AI and HTML. Then move into a visual tool like Figma only for the part that requires judgment.—KP

Build with Every

Every is a media company, a software company, and a consulting company—all run by a team that ships like an organization 10 times its size. If you’ve been wondering what working at the edge of AI looks like, we just opened up five new roles at Every:

• GTM engineer
• Head of finance vertical, consulting
• Head of learning and development
• Head of product marketing
• Head of social

Follow Every on X at @every and on LinkedIn.

We build AI tools for readers like you. Write brilliantly with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora. Dictate effortlessly with Monologue. Work on documents with AI agents using Proof.

For sponsorship opportunities, reach out to [email protected].