The Studio Bible

This is a big document on purpose — it's a reference, not a novel. You don't read it cover to cover. Start with Part A to get up and running, then dip into the rest whenever you hit something new. Every chapter is self-contained.

You've been given a working copy of a design studio that Aaron built and refined over many months — re-pointed at your accounts. It is not a blank tool you have to figure out. It already knows how to design websites, decks, dashboards, brand systems and social assets to a high standard, and it already has a quality-control process baked in. Your job isn't to build the engine. It's to learn to drive it.

You've used Claude (or ChatGPT) in a browser — you type, it types back. Claude Code is the same intelligence, but with hands. Instead of only writing answers into a chat box, it can open your files, change them, create new ones, run commands, and save its work — all inside one folder on your computer. That difference is the whole point.

There's a good chance Aaron (or whoever set up your laptop) has already installed this for you — if so, skip to Step 4 and just learn how to open it. If you're starting from a clean machine, here's the whole thing. Don't rush; you do this once.

Here's the screen you'll be looking at. It's simpler than it seems — really just a conversation that scrolls upward, with a box at the bottom where you type. Below is a mock-up of a real exchange, with the parts labelled.

The single biggest lever on quality is how you ask. Same engine, wildly different output. The difference is detail and context.

"Git" sounds technical and you'll never need most of it. But two ideas — commit and push — are worth understanding, because together they mean you can never really lose work or make an unfixable mistake. That peace of mind is what lets you move fast.

Claude starts each session blank — it doesn't remember yesterday unless you give it a memory. You already have one built in: a folder of tiny notes, each holding one fact, with an index that loads automatically at the start of every session. This is the single thing that turns Claude from a clever stranger into a colleague who knows your studio, your clients and your taste.

You'll use Claude in two places, and it's worth being clear which is which. Claude Code (the terminal) is where you build — it touches files. Claude.ai in your browser is where you think, draft, research and chat on the go — including a "Project" you can set up to mirror your studio's knowledge so it answers like it knows your world. Same intelligence; different hands.

Here's how to stand up a co-work Project on Claude.ai that mirrors your studio's "mythology" — its operating rules and memory — so it replies with the same context Claude Code has.

Beyond chatting, Claude.ai can run on a schedule and reach into your connected tools — so it can do small recurring jobs without you starting them. This is the "other place for basic automations" you mentioned. Keep the principle simple: automate the repetitive thinking, not the creative building. The building stays in Claude Code where you can see and steer it.

You don't need to build a design archive by hand — the studio has one. It's a gallery-first dashboard that scans every folder of work, screenshots each page / image / PDF into a thumbnail, and renders one scrollable page where each account is a collapsible section of cards. Click a card → the real file opens. It's a single static HTML page you open by double-clicking a .cmd file. No server, no upkeep.

Every serious visual output runs through a "board of directors" — sixteen named lenses, each a famous studio or designer, each asking one sharp question. It's not bureaucracy; it's a checklist that catches the things you'd otherwise ship and regret. Trigger it by typing /design-review or saying "run it through the board." A few directors are tuned for finance (faded below) — optional for a creative agency — but the core ten apply to almost everything you'll make.

Fast Lane — for quick internal work, ask for Tier 1 + the final gate only (6 directors, ship at 5/6). Use it when speed matters and stakes are low. For client-facing work, run the full board.

The studio is built in layers, from the quality gate at the bottom to finished video at the top. You don't operate these directly — they're how the commands work — but it's worth knowing they're there. teal = core, daily   blue = creative-agency essentials

These already exist in your repo. Commands are shortcuts you type (start with /) that run a whole workflow. Departments are folders the work is organised into. Skills are specialist know-how Claude loads on demand. You'll learn them by using them — these are the day-one ones.

Most of what the studio makes costs nothing — it's HTML, CSS and design work the model does directly. The only thing that costs money is generating photoreal images through a paid API. The whole discipline is: never pay for an image you can make for free, and when you do need a real photo, use a subscription you already pay for rather than a per-image API. Two rules cover it.

Aaron's routing matrix — which tool wins which job (adapt the platform names to your own subscriptions):

These are practices that took Aaron a long time to learn. Adopt them now and skip the painful version.

When a word trips you up, it's almost certainly here. None of these require any technical knowledge to use — they're just labels for things you already understand.

Everyone hits these in week one. Here's the fix for each — and the meta-answer: when in doubt, just ask Claude in plain English. "I'm confused about X, explain it simply" works for almost everything.