Corpus - what we were looking for
Check intuition that Nuprl proofs are close to human proofs
- tactic steps correspond approximately to inferences people would take
- steps at same level of granularity as human proof steps
Look for evidence of concrete proof strategies and communication patterns
- ideally, proof steps fitting the same role in a strategy are communicated in the same way