brew log.

Higher flow = less puck resistance. Beans losing CO₂ should push flow up at a fixed grind, so that was the hypothesis this chart was built to test. It is not showing up. Pooled over all 281 clean shots, with every bag-and-grind cohort centered on its own means so that neither level differences between bags nor a grind change mid-bag can pose as a trend, the slope is +0.04 ± 0.04 g/s per unit of ln(day + 1), which is not distinguishable from flat. Any real drift is smaller than the shot-to-shot scatter at this sample size. The individual bags also disagree with each other by more than noise allows, so one shared degradation shape is not a claim this data supports. These figures recompute on every build. Each curve is a logarithmic fit (y = a + b·ln(x + 1)) on clean shots only; it captures the fast-early, slowing degradation shape without overfitting the handful of data points per bag. Fits are computed per grind setting — bags brewed with multiple grinds get one curve per grind, distinguished by stroke style (solid, then dashed, then dotted in the order each grind was first used). The pale ribbon behind them is this bean's own history: the middle half of every clean shot it has produced across all of its bags, as a sliding quantile on days off roast. It describes where shots have landed and deliberately asserts no trend. Hover it for the shot and bag count. Solid dots are per-day averages of clean shots, hollow dashed rings are individual clean shots, and × marks shots excluded from the math (dial-in sweeps, prep mistakes, or process errors — hover for the reason). Shaded band is the rough espresso peak window for the selected bag's roast level — dark 7 – 14d, medium-dark 8 – 17d, medium 10 – 21d, light-medium 14 – 24d, light 18 – 28d, ultra light 21 – 42d. Darker roasts outgas faster and stale sooner, so their window opens and closes earlier. Bags with no roast level recorded get no band.