HN folks,

i built a plain TXT pack that turns 131 hard questions (math/physics/consciousness/AI alignment/etc) into a shared “effective-layer” interface: each item has scope, assumptions, and what would count as failure, plus prompt templates so LLMs run it consistently.

not claiming any problem is solved. goal is to stop endless debate loops caused by mismatched starting vocabulary.

MIT license, on GitHub, ~1.4k stars. happy to share a mini-index in comments if useful.