Primer: Prompting, Tokenization, and Input Parsing for Games
Lab design artifact — /lab
TL;DR: Prompt outcomes hinge on set theory and semantics—not rote arithmetic. Disambiguate scope, units, and partitions (singular vs plural; one word vs a list) to avoid gotcha prompts and yield stable, checkable answers.
What to look for
- Explicit units: define the token (e.g.
e), word form, and whether it is written once or per item - Partition awareness: Paths A/B/C produce different totals—different partitions, different answers
- Unambiguous quantifiers: use counts and constraints instead of colloquialisms
- Style attractors: avoid phrasing that nudges models into stock patterns; prefer neutral schema
This primer pairs with the Directive Hierarchy Framework—context and tier first; linguistic precision second.