Observable facts
Sources, excerpts and verification dates travel with each case.
Research standards
Evidence before conviction.
A useful case should help you continue, narrow the scope, test one mechanism, or walk away.
Implementation hypotheses
Possible implementations are labelled as analysis, never presented as recovered source code.
Original recommendations
Blueprints turn a direction into one bounded experiment with acceptance and stop criteria.
Unknown stays unknown
Missing revenue, team size, timelines and AI-use details are never replaced with invented values.
AI games, with evidence and limits.
Inclusion requires traceable developer statements or platform disclosures. Missing disclosure does not prove no AI use. Sources and necessary limitations remain public.