Companions that remember your style. Coaches that watch your last run. Briefings that catch you up the moment you log in. A shared memory that survives across every world — powered by the model you choose.
Each surface is a different shape of intelligence. They all read from the same context — your worlds, your sessions, your style — so the AI never starts from zero.
Persistent characters that remember your style, your inside jokes, and your last conversation. They don't start over every time you log in.
"Welcome back. Want to pick up where we left the chapter, or try something different first?"
A shared context layer that all surfaces read from. Your companion in one world knows what you learned in another.
Mentor in EDU references the moment you got stuck in Worlds last Tuesday.
A 90-second catch-up the moment you sign in. What your friends shipped, what the world moved on, what you missed.
Wren shipped chapter 04 of Riverbend. Cohort merged the new mechanics — review thread is hot.
Watches your last run and flags two or three patterns worth fixing. Specific. Actionable. Never patronizing.
You're losing 1.2s in the second turn. Cause: late braking. Drill 2 has the timing window.
Finds players you'll click with based on rhythm, style, and what you actually do, not just what you queue for.
3 lobby spots open with players who match your pace and play your style — joining in 18s.
Surfaces worlds, creators, and lessons you'll actually want. Less algorithmic feed; more "here's what fits."
Three worlds shipped this week that match your last build. Two have your friends in them.
Eternia's AI doesn't ask you to re-explain yourself. It reads your last session, your active worlds, and the friends you play with — then meets you where you are.
I'm stuck on the second turn — keep losing pace.
I watched your last run. You're braking 0.4s late into the apex. Try drill 2 in the Riverbend track — it has the exact timing window.
Can you queue it up?
Done. Loading in 5s. After the drill I'll mark whether the pattern stuck — if it didn't we'll try a different angle.
The intelligence layer is built around three concerns. Each one is independently swappable so the platform isn't owned by any single vendor.
The memory is yours. It lives with your account, not with any one provider. Switch models without losing what your AI knows about you.
The right model for the moment. Long story? Claude. Quick multimodal? Gemini. Sensitive? Local. The router picks; you can override.
Your keys, your residency, your call. Eternia never sees the conversation — we hold the memory index and let your model do the talking.