You figure out what you actually want.
The system asks the questions that surface your own meaning, then steps back.
How it works
How a thought becomes an intention.
An intention rarely arrives fully formed. You start with a thought, find its shape in conversation, name it when you recognise it, and commit when you are ready. Nothing is overwritten on the way, so what you meant last year sits beside what you mean now.
What this changes
Three things you notice.
You stay the author of your direction.
Suggestions stay provisional. Changing your mind is treated as authorship, never failure.
Your past selves remain part of you.
Earlier wordings are never deleted. When something old becomes relevant again, the system surfaces it.
Capture
Begin where the thought is.
No format. No template. Write what is in your head, raw and half-formed. Etcera receives without judgement and without restructuring you before you are ready.
Clarify
Find the shape of your own meaning.
The assistant asks questions that surface your meaning. Never questions that lead toward a predetermined answer. The output is wording you would recognise as yours.
Recognize
Name what was already forming.
You decide that something meaningful is present, and you name it. Etcera may notice an intention emerging, but it never decides on your behalf that one exists.
Commit
Act when you are ready, not when prompted.
Commitment is taking responsibility for carrying an intention forward, not a feature to unlock. It belongs to the intention itself, so you can change how you pursue it without breaking it.
Reflect
Update the shared understanding.
After action, the system records what changed. Earlier wordings are never deleted; the lineage of your becoming stays traversable. Releasing an intention ends its claim on your future without erasing its place in your story.
Your story, and where you are in it
Standard AI flattens you. It overwrites the past with the present and treats the latest version of your behaviour as the whole you.
Etcera keeps a story as a continuing thread rather than a final summary, so one story can hold that in 2024 you believed X and in 2026 you believe Y. Both are kept. Both are relevant. When a past intention becomes meaningful again, it surfaces without being prompted, and without overwriting where you are now.
What you see at any moment is your state: which stories are active, which intentions are guiding you, what feels possible right now. It is reconstructed from what you have shared and authored, never stored as a fixed version of you. It reflects where you are. It does not decide what you do next.
The longer argument
Intention Is All It Takes
Read the paper for the longer argument. Today's models often attend to patterns in data, with no intrinsic connection to what a person cares about. Binding a language model and a reasoning module to human intention is what closes that gap.
The five moves above trace what the paper calls Think, Do, Reflect. It makes the case for that loop, and for why one move answers two problems at once: AI alignment, and human agency. Read it in full, or see what Etcera is for.