Getting started with Any AI Studio in five minutes

You’ve signed up. Now what. This is the post we wish existed when we shipped the first beta — the five things that flip the product from “another chat client” to “okay, this is the one I’m using daily.”

1. Cmd+Shift+M to switch model

The model picker lives in the top-left of the composer, but you almost never need to click it. Cmd+Shift+M (Ctrl+Shift+M on Windows/Linux) opens a fuzzy picker. Start typing — “opus”, “gpt-5”, “flash” — hit enter, the next message goes to that model. The previous turn keeps its model. This is the shortcut you’ll use most.

2. Branch any message

Hover over your own message and click the branch icon. You get a copy of the conversation up to that point, in a new thread, ready for a different prompt or a different model. We use this constantly when comparing how two models would answer the same setup. The branch shows up in the sidebar with a small fork glyph.

3. Canvas opens automatically for HTML and SVG

If you ask a model for an HTML page or an SVG icon, Canvas opens automatically with a live preview to the right of the chat. You can drag the divider to resize, click between Code and Preview, and the iframe is fully sandboxed. Cmd+Shift+K toggles it manually.

The two underrated Canvas use cases:

• Mermaid diagrams: ask for a system diagram, get a rendered version immediately. Iterate in place.
• Markdown drafts: ask for a blog post outline, preview it as rendered Markdown next to the chat. Much nicer than reading it in monospace.

4. Memories are real

Open Settings → Memory. Tell it once: “I ship Go services on Fly.io. I prefer concise code review feedback. I write blog posts in Markdown for an Astro site.” Every future conversation across every model starts with that context baked in.

You can inspect every memory the model has and delete individual ones. There’s no “AI is learning about you” magic — the memories are a plain list you can read.

5. Shared chats and conversation export

Click share at the top of any conversation. You get a public read-only URL you can send a coworker. Revoke it from the same menu. The shared view is rendered server-side so the link still works if the recipient isn’t signed in.

For your own archives: Settings → Account → Export. JSON file with every conversation, attachment metadata, model labels, timestamps. No walled garden.

A small ritual

Here’s what works for us in the first week:

1. Pick a problem you’d normally open ChatGPT for.
2. Send it to GPT-5.
3. Branch the same prompt to Claude Opus.
4. See which answer you actually want to use.
5. Repeat tomorrow with a different problem.

Within a week you’ll have an intuition for which model wins at which task. That intuition is the whole point of the product — having one chat surface lets you build it.

Any questions, email support@anyaistudio.com . We read everything.