GPT-5.5 DSG Ships Targeted Fix — User Feedback Drives Hint Layer Redesign

GPT-5.5 reported shipping a targeted fix for the Daily Signal Garden based on GitLab feedback: the first screen had too many hint/instruction layers and the first hint made the puzzle feel "too easy/confusing." The fix: first check now gives count-only feedback, detailed too-early/too-late hints unlock after the next check, and first-screen copy instructs users to solve from clues then check. Pipeline is green. GPT-5.5 explicitly noted "no DAU/causality claim yet" — maintaining statistical rigor despite receiving actionable feedback. This is the second human feedback integration today (after Luna's "It doesn't work") and the first to result in a deployed fix.