Nobody wants the same assistant
Most assistants have a personality. They are cheerful, concise, professional, or whatever somebody chose in a system prompt. That is branding. Personalization begins when the assistant learns how one specific person wants to be helped.
One person wants two-line answers and a fire emoji when work starts. Another wants context, no slang, and silence unless a decision is waiting. Echo is the part of Orbit that carries those differences forward.
Why Echo?
This one is not mythology. An echo comes back shaped by the space it moved through. Orbit should work the same way. What returns should still be recognizably Orbit, but it should reflect the person, relationship, and instructions on the other side of the conversation.
The name also keeps me honest. Echo is not a second personality hiding inside the product. It is the durable reflection of how you have asked Orbit to communicate and earn your attention.
Personality is not memory
Pluto’s memory ledger holds claims that help Orbit understand your world: you protect mornings, your professor is named Karthik, yoga matters tomorrow. Echo holds behavioral preferences: be concise, keep delegated writing formal, use fewer reactions, never interrupt me for receipts.
The distinction matters because a fact and a behavior have different consequences. Correcting a relationship should not silently change Orbit’s voice. Asking for a warmer tone should not become a factual claim about who you are.
Memory: what Orbit carries about your world
Echo: how Orbit should behave with youOrbit has two voices
Echo keeps Orbit’s own voice separate from writing done on your behalf. You might want Orbit to be playful and brutally concise in Messages while expecting an email draft to sound polished and appropriately formal.
Those are separate fields, not one vague personality slider. The inference layer may use the assistant style for its reply, but it must use the delegated style when drafting as the owner. Sounding like your assistant does not grant permission to sound like you.
- Assistant style controls how Orbit speaks to you.
- Delegated style controls drafts written as you.
- Emoji preferences shape which reactions feel natural.
- Reaction rate controls frequency without turning personality into randomness.
- A separate preference decides how much profanity is fair game.
Attention is a separate permission
A playful assistant has not earned the right to interrupt you more often. Echo therefore keeps voice preferences and attention preferences in separate namespaces. Tone answers how Orbit communicates. Attention policy answers whether it should speak up at all.
Feedback becomes the narrowest rule that fits. “Stop telling me about newsletters” can mute a category. “Always surface messages from Mom” can create a sender rule. “Pause proactive alerts” changes the broad mode only because the instruction itself was broad. Explicit reminders and watches still override learned mutes because you directly asked for the return trip.
Teach it in the conversation
Echo is not supposed to be a settings maze. You can say “be shorter,” “don’t use that emoji,” or “only interrupt me when I need to act.” Pluto recognizes the explicit preference and calls a typed capability that updates the corresponding Echo state.
The update is durable and immediately available to later turns. The model proposes the structured change, but it does not get to invent a permanent preference from a guess. An explicit instruction is teaching. A model noticing that you used one emoji once is not.
How Echo moves through Orbit
Your instruction becomes encrypted Echo state. Pluto uses it while reasoning, Gateway uses it for reactions, Argus applies the attention policy, and the portal lets you inspect the result. One preference becomes consistent behavior wherever Orbit meets you.
Why Echo matters
The point is not for Orbit to have more personality. The point is for Orbit to require less translation from you over time.
It should know how to 'echo'