Think inside your AI world.
The authored turn: where the turn-based loop ends
A turn-based loop is the familiar one: you prompt, the model works until it judges the task done or needs more, and you read the result. Useful, and honest about its bottleneck — you. The authored turn keeps that shape and moves the ending one step: to a ratification gesture, with the decisions you’ve settled already in front of the model.
In the turn-based loop, the hand-back is the last node: the model returns something it judges complete, and acceptance is yours. The authored turn changes two things and nothing else — before the turn, the ratified decisions that bear on the work are injected into context; after it, the loop ends not at “the model believes this is done” but at a point where it surfaces anywhere the work touched or diverged from settled canon, and you confirm or correct. Unl holds the decisions; you keep authority over the judgments.
How does the turn-based loop end today?
Fairly described: a turn-based loop is triggered by your prompt and stops when the model judges the task complete or judges it needs more context. It’s the most controllable of the four shapes because you read every hand-back and write the next prompt. The quality practices around it — tests, a fresh-context reviewer, a review command — are sound.
The ending, though, is the model’s own read of “done,” and its context is whatever happened to be in the window. If a decision you settled last week bears on this turn and isn’t in the window, the turn can quietly re-derive it — and you catch that on the read, if you catch it at all.
What is the authored turn?
The authored turn injects the ratified decisions that bear on the work before the model acts, so the turn starts on your rails rather than the model’s defaults. Then it moves the ending: the loop closes at a ratification gesture. The model articulates where the work touched or diverged from settled canon, and you confirm it as new canon or correct it. The check you hand off is verification against tests; the check you keep is authority over judgments.
The model still does the reasoning at full speed. What changes is that a settled decision is present at the moment it matters, and the close of the turn is an explicit human act rather than a machine’s self-assessment.
Where does the ratification gesture live in a turn?
Per turn, at the hand-back. That’s the natural home for the gesture in this shape: the human is already reading the output, so the loop asks for one more thing — a confirm or a correction on anything that touched settled canon. It costs a gesture, not a re-brief, because the divergence is surfaced rather than buried in a diff you have to reconstruct.
The authored turn ends at a ratification gesture, not at “the model believes it’s done” — judgment present at the read.
Reads through Unl arrive with measured context — in the presence of the decisions you’ve already settled. The reach lane is live: one box, paste anything. If it speaks MCP, Unl can reach it. Readings arrive unprompted, the data beside the criterion; Unl is a courier, not a warehouse, and keeps only your keys and the frame.
Read further
Questions people ask
What is an authored turn?
A turn-based loop with two changes: the decisions you’ve ratified are injected into context before the model acts, and the turn ends at a ratification gesture rather than the model’s own judgement of “done.” The model surfaces where the work touched or diverged from settled canon, and you confirm it as canon or correct it — a gesture, not a re-brief.
How is this different from just reviewing the output?
Reviewing catches problems by eye, after the fact, without your settled decisions in front of you. The authored turn puts the ratified decision in context before the work and flags divergence from it at the hand-back — so you’re confirming against what you decided, not re-reading to guess whether the model stayed on your rails.
Does the model still work autonomously within the turn?
Yes — it reasons and drafts at full speed. Authorship is what stays with you: the model proposes, and canon changes only when you make the gesture. Reads through Unl arrive measured against your settled decisions; the reach lane is live and open: if it speaks MCP, Unl can reach it.
What this is
Think inside your AI world — you stay in command
Unlimitless (Unl to friends) holds what you've settled, reads what your tools are showing, and catches what's changed out in the world — and hands your AI whatever bears on the work, the moment it's needed, without you asking. The right thing, in front of the model, unprompted, with you in command of the call. So you keep moving toward what you set out to build, on top of everything you've already decided.
It plugs into Claude, Claude Code, ChatGPT and Cursor as an MCP connector. Quick to connect, in a couple of steps.
Unlimitless is open now to invited Alpha. Apply for the Beta waitlist to come in ahead of the full launch:
Alpha is invite-only · free at launch.