Past the sliding glass door of a mid-century house in West Adams, Los Angeles, through a courtyard where a guava tree drops fruit nobody picks up, there is a back studio with three monitors and the kind of natural light that photographers spend whole careers chasing. Galen Oakes is not taking photographs anymore. He is thirty-nine years old, an artist who taught himself to code, and he has been sitting in this room since September, seven in the morning to one at night, building an operating system for the way small businesses will run when AI finishes rewriting the rules.

He grew up in a cabin in the redwoods of Northern California. Westport, population sixty, the last village before the Lost Coast, the only stretch of American shoreline that was too steep and too wild for the highway builders to touch. His grandfather engineered nuclear submarines and came home with briefcases handcuffed to his wrists. His father was an environmental artist who took him to Macworld and bought a Centris 650 to publish a book called Sculpting with the Environment. His mother was a teacher who put a camera in his hands at sixteen, a Sony Cyber-Shot, this silver brick with a tiny screen, and he took it everywhere. He pirated Photoshop in middle school, aced an advertising final in community college without studying, got an incomplete in another class, dropped out. Trimmed pot for a year to save money. Went to Bali. Went to Burning Man. Moved to San Francisco at twenty-one, before Instagram existed, before Uber, and spent the next fifteen years shooting festivals for Red Bull, directing campaigns for Nike and Salesforce, helping build The Village, the premier tech venue in the city. Five companies. Every one of them was somebody else's dream. Every one of them ended the same way.

Oakes has ADHD, which he will tell you is a blessing and a curse in the same sentence and mean both parts completely. The blessing is that his mind holds a whole portfolio at once, sees the thread connecting a restaurant OS to a stage product to a knowledge system, and follows that thread while everyone else is still opening their first tab. The curse is that the world was never built for the way he thinks. School was not built for it. The five partnerships that ate his twenties and thirties were not built for it. The traditional path from idea to company, the one that requires money and developers and project managers and eighteen months of runway before you ship anything, was not built for it. Then AI arrived. And the skill he had carried his whole life, the ability to see something clearly in his mind and describe it into existence with words, turned out to be the exact skill you need to build with it.

That is what he is building toward, late at night, when the screens are the only light and the guava tree outside is just a silhouette against the street. Help creative people have agency. Give them the tools he never had. Build the thing he needed when he was twenty-five and talented and completely dependent on someone else's timeline, someone else's money, someone else's willingness to keep going.

Galen Oakes has ADHD. His mind sees patterns across the whole portfolio simultaneously, connections between a restaurant OS and a stage product and a knowledge system in the same breath. It also means that if he does not write it down, it disappears. The living strategy docs were not a slide exercise. They were a survival mechanism.

It started with Cursor. Most people use it to write code. Oakes opened it and started writing words. “I was using Cursor as a text editing platform. Not for code. I was saving all meetings inside Cursor, all research, all plans. For synthesizing and context engineering and coming up with plans.” Meeting transcripts, research, strategy documents, organizational maps. All of it living inside one editor, all of it feeding into the AI. Cursor became the playground for everything that followed. By June 2025, he had built the first version of all five documents inside Cursor. He authored 100 percent of his company's knowledge base this way, not by coding, but by curating the context that would allow an AI to understand a business the way a chief of staff does after three years of sitting in every meeting.

He built five living documents. Strategy. Roadmap. Knowledge. Stakeholders. Results. Each one feeds from nine data sources: email, meetings, Slack, Telegram, iMessage, WhatsApp, entity interactions, deep context, and canvas documents. Every day, those sources produce new information. Every day, the documents update themselves. A planning model reads the incoming data and proposes specific edits. An execution model applies them surgically. The philosophy is amend, not replace. The documents are continuous prose that grows and evolves. They never truncate. They never summarize away what came before.

“Every business will only be as powerful as their knowledge base,” he told a partner in July 2025, “and their knowledge base will only be as powerful as the way that they've engineered the context. Meetings are context. Slack channels are context. Research is context. Previous documents are context. All of these things work together to give you superpowers.”

By early 2026, every major AI agent framework had converged on exactly this pattern. OpenClaw, Claude Code, every serious agent system runs on text files that tell the AI how to think and what to remember. The entire industry landed on documents in folders as the operating layer for AI. Oakes had been building this in production for a year before most of these frameworks shipped. But he took it further. Instead of static files sitting on a laptop, he moved the entire system into a cloud database where any agent, anywhere, can read from and write to the same shared brain. The documents are not files anymore. They are living rows in a database that search by meaning, not just keywords, and rewrite themselves as the business changes.

Then the line that captures the entire intellectual project: “Global styles for intelligence.” If you have built websites, you know what global styles are. One file controls the fonts, colors, and spacing across every page. Change it once, change everything. Oakes asked what would happen if you did the same thing with company knowledge. One set of living documents. Every agent in the system reads from them. Update the strategy document and every agent, from the customer service bot to the marketing writer to the project planner, adjusts its behavior.

On June 11, 2025, he committed his Knowledge Curation Thesis to GitHub. Eight days later, on June 19, Tobi Lutke, the CEO of Shopify, posted a memo that popularized the term “context engineering.” The industry adopted it within weeks. Oakes did not have the term. He had the practice. The timestamp is in the git log.