Question ranking
What direction does this Yawn serve?
Make the next question useful, inspectable, and correctable without mistaking preference for truth.
proposed · Synthesis of Dave's September 11 design request; open to correction.
What is a purpose vector?What makes one question worth asking before another?
Its relevance to a chosen purpose, the uncertainty it could resolve, the decisions or questions depending on it, and the effort, risk, and permissions involved. A ranking always embodies assumptions and priorities; it cannot promise a view without bias.
Why is this question here?
Governing inquiry · fixed first
This question orients the model. Reading activity cannot displace it.
The remaining questions follow the selected ordering mode.
What actually determines the order on these pages?
Implemented presentation policy
The governing question stays first. Authored and manual modes retain their explicit order. Learned mode puts your protected questions first, then compares log₂(1 + recorded openings) + 3 × relevance vote + a capped count of distinct dependent questions. Ties preserve authored order.
Why is this question here?
Authored order
No numerical score determines this position in the current mode.
Placed before math by the source order.
How is this answer supported?
Dependency contribution is at most 3 and uses only declared links among questions in this model. No private or unavailable graph is read. Openings contribute only after opt-in. Agreement and notes never enter the score.
Purpose influences your explicit Keep in view and relevance choices here. The runtime does not infer a numerical purpose vector, expected information gain, or probability from your words. Changes wait for an explicit Apply learned order or a new visit so the page stays still while you read.
What happened to the earlier math?
The earlier question scorer combined expected-information-gain, action-value and risk-reduction proxies, minus asking cost and friction. The dual-lattice implementation linked skills and outcomes with confidence weights. These are useful engineering precedents, not established mathematical laws of cognition.
Why is this question here?
Authored order
No numerical score determines this position in the current mode.
Placed before navier-stokes by the source order.
How is this answer supported?
Open PR 124 proposes Priority = information gain × endorsed-purpose relevance × consequence × decision impact / max(0.05, burden + risk + delay), after hard gates. It is an unmerged proposal for tests and views, not the implemented reader order and not a probability. Unknown purpose and invalid inputs still need careful treatment.
What would make the next question better?
Record whether it changed understanding, exposed a constraint, or enabled a useful move. Compare alternatives and failed predictions. A repeated click measures attention; a protected commitment records a choice. Neither proves an answer true.
Why is this question here?
Authored order
No numerical score determines this position in the current mode.
Last question in the current order.
