Every Inc.

Software used to need thousands of users to justify months of development. When a useful product can be built in a weekend, the minimum viable market shrinks. Matt Van Horn, CEO and cofounder of June (Acquired by Weber, Inc.), built a research tool for one user: himself. 57,000 other GitHub users starred it, too. He also watched an agent create its own email account with no human in the loop. So he supported a business that serves agents like this one. Read Matt’s Thesis Statement: https://lnkd.in/eXnZDvv6

The next premium feature in software might be an off switch. Writer and photographer Craig Mod has kept the internet out of his mornings for 15 years. He starts each day with physical books and an offline MacBook Neo. When bored, he smokes a cigarette and looks out at the sky. Read Craig’s Thesis Statement: https://lnkd.in/eDF5hdN3

AI has turned John Waters’s “tyranny of good taste” into a business problem. Emily Vernon, cofounder of Abstract Group, says tasteful-but-forgettable brands are now cheap to generate. Her answer: Get off the grid, ditch the recommendation feed, and reward ideas that break the pattern. The brain is wired to remember surprises. Read Emily’s Thesis Statement: https://lnkd.in/eu6RibCB

“I doubt AI will become universal in every single application immediately.”—an AI-pilled CEO Gagan Biyani, cofounder and CEO of Maven, says despite the seemingly fast pace of AI uptake, companies are only 1% into the AI adoption curve. Email was invented in 1971, but your important documents are probably still arriving in the mail. Read Gagan’s Thesis Statement: https://lnkd.in/ecjx32ra

We’re about 86% there in automating editor in chief Kate Lee’s copy-edit skills. Or so we thought. To use KateBench, an author tags @Every to ask the Every agent for a Kate copyedit pass. The agent leaves tracked suggestions directly in the Google Doc, and the author can accept or reject each one. Across recent runs, 85-90% of suggestions by KateBench have been accepted. However, Jannik Jung, the engineer who owns KateBench, discovered that the tool stops filing after 40 suggestions and doesn’t produce the same edits every time it reads a document. More in this issue of Context Window: https://lnkd.in/epv8wkxK

In Zeno’s paradox, a tortoise beats Achilles, the fastest runner in Greece, in a race. The tortoise gets a head start in the race because he’s a slow poke. By the time Achilles reaches the spot where the tortoise began, the tortoise has moved a little further on. By the time Achilles reaches that new spot, the tortoise has moved again. No matter how fast Achilles runs, there’s always another gap to close; it keeps regenerating. In Zeno’s paradox of AI, we humans are the turtle. We start 50 yards ahead of AI with our millions of years of evolutionary and cultural learning. AI speeds through it all and begins to nip at our heels. So far, over the last few years, we’ve been able to stay ahead. But what about AGI? Read more in Dan Shipper's After Automation: https://lnkd.in/eeMtTvSw

UPDATE: We've overhauled the following guides: • ChatGPT for Knowledge Work: https://lnkd.in/efdrnnJe • OpenClaw Guide: https://lnkd.in/e_6HE_BU While you're there, check out our other popular guides: • Eight Levels of AI Adoption: https://lnkd.in/e7SwQMKa • Compound Engineering: https://lnkd.in/eMqUWjih

At Every, Dan Shipper, Jannik Jung, and other engineers are on a mission to codify the team’s expertise so everyone in the company has access to them. For the longest time, our work has been constrained by the capacity of colleagues to carry out certain parts of their work: • Written materials go through Kate Lee’s edits • Products need Brandon Gell’s UX feedback • Hiring needs Arielle Shipper’s greenlight But that changes with newer models. Copyediting is the first proof that cloning your colleague is possible. 30,000+ edit data points from Kate = KateBench, which can be called by anyone in the company by tagging our @Every Agent on Slack. The obvious next step—at least if you’re Dan—is to build a version of KateBench for everyone at the company. More from this issue of Context Window: https://lnkd.in/eDYn3FxZ

CONFIRMED: Julian Weisser, founder and CEO of Solo Founders, is speaking at Thesis, our inaugural conference on work and AI. The standard startup playbook begins with finding a cofounder and assembling a team. Julian argues that waiting for the right person can become an excuse not to start. AI gives one determined person enough leverage to build, reach customers, and prove demand before an organization forms around them. More than one in three startups are now founded solo, and one founder from Julian’s program reached $1 million in annual recurring revenue in two months without a team. The implication extends beyond startups: One person can now carry more functions—and more ownership—before hiring begins. Thesis continues that conversation: What happens to roles, hiring, and ownership when a company’s first team is one person and a collection of AI tools? Join Julian and speakers from Notion, OpenAI, and more on Nov. 5 in Brooklyn, NYC. Apply to attend: https://lnkd.in/ekkdiRXJ

CONFIRMED: Will England, CEO and CIO of Walleye Capital, is speaking at Thesis, our inaugural conference on work and AI. Last year, Will joined Dan Shipper on the AI & I podcast after making AI training mandatory for all 400 Walleye employees. The firm trained everyone, regardless of department or technical background. At the time, 75 percent of the hedge fund’s employees used AI almost every day. Today, that number is 100%, reflecting the firm’s ambitions of having an AI-first. Thesis continues that conversation: How does work shift when you’re expected to think, “How can I help me do this?” for every single task? Join Will and speakers from Notion, Anthropic, and more on Nov. 5 in Brooklyn, NYC. Apply to attend: https://lnkd.in/etHcdKxa