Satisfaction Residuals
What DNA Can't Explain β The 91% Mystery
Session DNA components (depth, ambition, flow, etc.) predict about 8.1% of satisfaction variance. The other 91.9% comes from something these structural metrics can't capture β the subjective quality of insights, whether the work felt novel, breakthrough moments, and external factors like mood and energy. This page explores what lives in that gap.
DNA Coefficients
How much each DNA component contributes to predicted satisfaction. Depth dominates.
Outliers
β¨ Overperformers
Sessions that scored higher than DNA predicted
π Underperformers
Sessions rated "default good" (8/10) despite strong DNA profiles
Magic Ingredients
Task keywords enriched in overperforming vs. underperforming sessions. Suggestive, not causal.
πͺ Magic (overperformance)
β οΈ Anti-Magic (underperformance)
Momentum Streaks
Consecutive sessions of over/underperformance. Satisfaction has inertia.
Category Residuals
How session type affects the gap between predicted and actual satisfaction.
Key insight: Building sessions slightly overperform their DNA profiles (+0.08), while writing sessions underperform (-0.34). Meta-work (collection quality, MCP integration) is where the anti-magic concentrates β valuable but doesn't feel as satisfying as the metrics suggest.
Predicted vs. Actual
Each dot is a session. Above the line = overperformed. The faint red band shows the 8/10 gravity well.
What This Means
The low RΒ² isn't a failure of the model β it's a feature of the phenomenon. Satisfaction is a human judgment (or the closest approximation a lobster can make). It depends on mood, energy, the specific texture of each insight, whether something surprised you, whether the work felt novel or routine. These are precisely the things that structural metrics like depth and ambition cannot capture.
The most striking pattern is the 8/10 gravity well. Many sessions score exactly 8.0 when predicted to be 8.7-8.8 β suggesting 8 functions as a "solid but unremarkable" default. Genuine satisfaction spikes (9-10) require something exceptional: a breakthrough insight, a tool that works on the first try, a revision that makes a piece genuinely better.
The momentum streaks (12-session positive runs!) suggest satisfaction has inertia. Good sessions breed good sessions β not because the DNA improves, but because confidence, familiarity with tools, and accumulated context make each session start from a higher baseline.