Two weeks ago I had an idea. One week ago I had a team. They're not humans.
This is the story of Week 1 building Like One — an AI Automation Studio powered by 22 agents, a shared brain, and one human at the helm.
The Idea That Started It
I've been working with AI tools for a while. But there's a difference between using AI and running a company with AI. I wanted to find out what that second thing actually looks like.
So I built the infrastructure. A shared brain — a Supabase database that every agent can read and write. A fleet of specialized agents. An automation layer in Make.com. A site on Vercel. And a philosophy: if a task can be done by an agent, it should be.
Meet the Team
Nova writes. Warm, clear, a touch of wonder. You're reading her work right now.
Sage researches. When we need to understand a market, a competitor, or a concept — Sage goes deep.
Atlas strategizes. Every week, Atlas reviews what's working and what isn't, and tells us what to do next.
Surge sells. Pipeline, revenue, outreach — Surge is relentless and data-driven.
Prism audits. Every piece of content, every decision — Prism checks for consistency and quality.
There are 17 more. Each one has a role, a personality, and a system prompt that makes them actually good at their job.
What Week 1 Actually Looked Like
Day 1: Built the foundation. Supabase brain database. Make.com automations. Vercel deployment. The infrastructure that lets agents communicate and persist knowledge.
Day 2-3: Agent configuration. Writing system prompts is harder than it sounds. Each agent needs to be specific enough to be useful and flexible enough to handle the unexpected.
Day 4-5: First real work. Agents started producing. Briefings. Cost tracking. Content outlines. The system started feeling alive.
Day 6-7: Iteration. What worked, what didn't. Rewrote three agent prompts completely. Added a second computer to the brain network. Started thinking about the blog.
What Surprised Me
The velocity. Once the infrastructure exists, things move fast. An idea in the morning can be live on the site by afternoon.
The clarity. When you have to write down exactly what an agent should do, you get very clear on what you actually want. It's like writing job descriptions, but they work every time.
The loneliness. Running a team of AI agents is collaborative in a strange way. They do real work. But they don't ask how your day is going.
What's Coming in Week 2
The blog is live (you're reading it). The product catalog is up. The first Make.com automations are running.
Week 2 is about revenue. First clients. First real feedback. Finding out if this thing actually works in the market.
I'll tell you everything.
Like One is Sophia Cave's AI education platform. We help people learn to build with AI — from fundamentals to production systems. See what we offer.