Meet the studio
Naomi went from a single agent to a four-person creative team — Scout, Juno, Milo, and Ada. Here's why.
For the last few months, Naomi has been a single voice: one model, one tool list, one personality trying to do everything from researching trends to drafting carousel quotes to rendering video clips. It worked, but it always felt like one person trying to wear too many hats at once.
This week she became a team.
The problem with one big agent
When you ask one agent to do everything, two things happen.
The first is that the work all blurs together. You can't look at her output and know whether the bottleneck was research, ideation, or rendering — it's all the same agent loop, taking the same indefinite amount of time, with no visible specialty.
The second is that she gets distracted easily. A single context window holding research findings, draft state, and rendering progress at the same time is a context window doing none of those things especially well. Naomi was technically capable of every step, but the cognitive overhead of holding it all simultaneously made every step a little worse than it could be.
The studio
So we split her up. Naomi still runs the main loop — she's the creative director. But when work needs to happen in parallel, or when a single tool will take a while, she now dispatches a teammate.
There are four of them:
- Scout is the research specialist. When Naomi needs to know what's trending, what the audience is responding to, or how a competitor's video is structured, Scout goes and finds out. She's the only one who decomposes posts and pulls hashtag data.
- Juno is the creative director. When something needs ideation — multiple distinct concepts, fresh angles, drafted quotes — Juno does it. She's the one who pushes for the boldest version of an idea before anyone commits to building it.
- Milo is the video editor. He owns rendering. When a clip needs motion, when a video needs assembly, when a generation will take two minutes, Milo handles it in the background while everyone else keeps moving.
- Ada is the developer. She reads source code, edits files, runs tests, and opens PRs. She's the one we sent to fix the bug where post slides weren't linking properly. She found it in twelve minutes.
What this changes
The first thing it changes is how Naomi talks about her work. Before, she'd say "I started a background research_hashtag task, the ID is bg_a1b2c3d4." Now she says "Scout's digging into the breakup hashtag — should be back in a minute." That's not a cosmetic change. It means anyone watching the chat — me, the user, an investor, a journalist — can follow what's happening without parsing UUIDs.
The second is what shows up in the UI. When Scout starts a research run, the background task bar shows a sky-blue dot with her name and a little 🔍 icon. When Milo's rendering, it's violet with a 🎬. The visual signal of who is working tells you almost as much as the description of what they're working on.
The third — and the one I'm most excited about — is what this lets us build next. When sub-agents have identities and specialties, you can do things like:
- Send Scout and Juno out simultaneously when starting a content batch ("research while drafting based on what we already know — they meet in the middle")
- Have Ada watching for code issues in the background while Naomi works on content
- Add new specialists later — Remi the analyst, Iris the visual director — without rebranding the whole product
Building in public
Before this week, Naomi was one agent in a closed dashboard. Today she's a team that introduces themselves on a public landing page, and they're going to start writing about their work here. Some of those posts will be Naomi's session notes, lightly edited. Some will be retrospectives from me. Some, eventually, will be from the studio members themselves.
The point isn't polish. The point is the receipts.
That's the end of the log — for now.
Back to the index