Questions, answered

Irenaverse is invite-only. Someone you trust — the parent of one of your kid's friends, a relative, the family who built this — sends a one-use link. You open it, your kid picks a display name and a passphrase, and they're in. No email address required for kids.

Designed around a 12-year-old. It works well from about 9 up to early teens; a younger kid will need a grown-up sitting next to them at first; an older kid will sail through the early lessons faster but the later ones (story mode, functions-as-values) still bite. There's no age check or restriction.

Not for invitees. Irenaverse is hosted on a server the founder already pays for, and the AI buddy has a per-kid daily cap that keeps costs negligible. There's no plan, no upgrade, no card on file.

Then they stop — there's no streak guilt, no notifications chasing them back. Their account just sits there. We'd rather have a kid who's excited the few times they open it than one who's nagged into daily use.

Real text-based code. Kids write Irene — a kid-friendly language that reads almost identically to Python. By lesson 7 they're writing functions; by lesson 21 they're juggling records, scenes and event handlers. The skills carry straight over to Python, JavaScript or any text-based language they meet next.

On a Chromebook, yes — anywhere Chrome or Firefox runs. On a tablet it'll display but the editor really wants a keyboard; an iPad with a keyboard case works fine, an iPad in your hands does not. The whole thing runs in the browser, no install.

A typical lesson is 10–25 minutes from first click to a playable result. There are 21 lessons in the main arc. Past that, kids drop into free-play with the editor and the bundled game templates (snake, platformer, asteroids, flappy, a few more) and tend to lose afternoons.

Not externally — sharing-out isn't built yet, by choice. Inside the invited circle, an admin can see what a kid has made; passing a game on to a sibling or friend in the same invite circle is on the roadmap. Public publishing isn't planned.

No — and that's the central design choice. Ren has three modes: Build (hints first, code only when asked, one or two lines), Learn (Socratic — questions only, never writes code), and Walkthrough (proposes a plan and walks one step at a time, checking the kid's code between each). The default is Build. "Just do it for me" isn't a button.

Ren sees your kid's current code and the chat history of the current project (truncated to the last 10 turns). Every exchange — the prompt, Ren's reply, token cost — is saved to a database row your admin can audit. There's a per-kid daily token cap (100,000 default; the admin can lower it or pause Ren entirely with one click).

Anthropic's safety training on Claude is the only filter between your kid and the model. There isn't an extra moderation layer in the middle today — see the How Ren actually works section on For grown-ups for the honest detail. For a wider circle a stricter posture is the obvious next step; ask whoever invited you to enable it.

A display name, a username, and a passphrase — no email address. Their code, their AI chats with Ren, their lesson progress. That's it. No analytics tracker, no advertising, no third-party SDKs. The data sits on one small server in New Zealand alongside about a dozen other family projects of the founder's.

The kid themselves and the admin (the adult who invited them, or the family who set this up). Nobody else. There's no public profile, no discovery, no leaderboard. If you want your account and everything in it deleted, ask the admin and it's a single command.