> The three classical ways a [[Definition|definition]] pins a term down
> - by [[Ostension|pointing]] at an example,
> - by [[Extension|listing the cases]] it covers,
> - or by stating the written rule that generates them ([[Intention|intension]]).
They are three rungs of a climb, each carrying further than the last: a written list outlasts the person who would otherwise be pointed to, and a rule answers cases the list never recorded.
^definition
Example:
> Take "dog":
> - you can **point** at one - *"that, and that, are dogs"* ([[Ostension|ostension]])
> - **list the cases** - *every dog there is* ([[Extension|extension]])
> - or **state the rule** - *"a domesticated canid that barks"* ([[Intention|intension]]).
^lens
In an organisation the rungs are usually climbed in that order.
1. First the knowledge is [[Ostension|ostensive]] - Sarah decides case by case, nothing written.
2. Then the decisions pile up into a record - a [[Decision log|log]], a ticket history, past examples: the [[Extension|extension]], written but not yet abstracted.
3. Only later does someone distil the rule - the [[Intention|intension]].
Recording what happened is cheap; the [[Distillation pattern|distillation]] that abstracts the rule is the expensive step, so organisations are usually **case-rich and rule-poor**, sitting on the middle rung.
Each rung buys what the one below cannot.
- The list is **written**, so unlike pointing it survives the person and can be handed over.
- The rule is **generative**: it answers cases the list does not contain.
A complete, finite list would make the rule redundant - you would just look the case up - but organisational knowledge is open-ended, so the list is never complete, and the rule is what covers the cases nobody has seen yet. That is why a fresh question - or anything an [[AI agent|agent]] is asked - needs the rule: the answer is not already in the record.