What Sophie reads from
390
GTM frameworks
784
Research papers
750
Operator essays
41,690
Citation links
297
Caugia helpdesk entries
Helpdesk corpus last synced: 18 Jun 2026, 20:58
Weekly independent grader
Every Wednesday, an independent Anthropic Haiku 4.5 model grades a 30-case sample of Sophie's answers against a public rubric. Scores publish to /admin/sophie-quality and the lowest cases trigger calibration.
Rubric: archetype precision (25) + framework relevance (30) + coverage (20) + coherence (15) + honesty (10) = 100
Sophie answers in any language you write to her in. Native UI + assessment + corpus in 6 (English, French, German, Spanish, Polish, Dutch) — other languages get fluent answers with English spelling preserved for framework names.