Kid mode πŸ§’

This page is Anatomy of a Jam β€” Window Picker. It is one of the projects Brandon built and put online. The grown-up details are below, tap Grown-up mode to read them the normal way.

PHISH Β· 2025 RIVIERA MAYA RUN Β· STEM-FIRST ANALYSIS

Anatomy of a Jam β€” Window Picker

Stems separated 2026-06-11. This pass scores every second of musical development across the run's five complete-stem jams, then picks the single best 4.5-minute window to analyze first.

The winner: Tweezer Β· 7:45–12:15

Highest development score of the run. One window, analyzed deeply, before moving to the next.

C
tonal center
min pent
lead mode (maj-key bed)
106
median BPM
86
tempo shifts
7
section onsets
0.772
window score

Read the ixi-style close analysis of this window, the C pedal floor, the held-G recolor, and the major-3rd flash, all grounded in the stems.

Why this window

Development curve across the full 15:30 jam. Shaded band = the chosen window. Lighter bands = runner-up windows. Toggle the drivers.

Inside the window

Loudness arc (window-local time) with the seven section onsets marked. These are the editor's cut points.

Mode / key timeline

Windowed tonal reading. The detectors land on the flat side, which is the point: a C-centered bed played through a blues scale.

Window timeReadingConfRunner-up
0:00–1:28 C MinPentatonic0.64 Bb MajPentatonic (0.69)
1:28–2:56 C MinPentatonic0.64 Bb MajPentatonic (0.69)
2:56–4:24 C MinPentatonic0.64 Bb MajPentatonic (0.69)

The read

The pick is data-driven, not a guess. Across the five complete-stem jams of the 2025 Riviera Maya run, the picker scored each second of music for energy, spectral novelty, harmonic motion, and instrument interplay, then slid a 4.5-minute window across the whole jam. Tweezer's window from 7:45 to 12:15 won by a clear margin, with an energy z-score of +1.16 and a novelty z-score of +1.01, both far above every other candidate. That is the part of the jam where the band is doing the most.

The window is a groove-locked Type II peak, not a chord excursion. The harmonic-flux reading sat slightly negative inside the chosen window, which means the tonal center barely moves while everything else intensifies. The mode detector agrees, holding C minor pentatonic across all three sub-segments. The whole window lives on one tonal floor and builds its drama through rhythm, density, and dynamics rather than through changing chords.

The bright-key, bluesy-lead tension is the real color. The two analyzers disagree on purpose. The global key detector reads C major at 0.74 confidence, while the windowed mode detector reads C minor pentatonic with B-flat major pentatonic close behind. Both are flat-side readings of the same thing. The band is centered on C with the flat seventh and flat third sitting in the lead, the classic Phish move where a major-key platform gets played through a blues scale. Take the major reading as the harmonic bed and the minor-pentatonic reading as what Trey is actually soloing over the top of it.

The tempo is alive. Median tempo lands near 105 BPM, but the analyzer flagged 86 local tempo shifts inside four and a half minutes. This is not a click-track jam. The pocket breathes, pushes, and pulls, which is exactly what makes a Type II peak feel like it is being discovered in real time rather than executed.

Seven section onsets give the editor the cut points. The jam_analyzer marked seven novelty spikes inside the window. Those are the natural seams where the texture resets, and they are the timestamps to hang narration on when this becomes a full anatomy-of-a-jam video.

Next in queue, analyzed: Golden Age Β· 6:30-11:00

Runner-up window of the run. A brighter, more resolved counterpart to the Tweezer peak.

F
tonal center
Ionian
mode (opens Mixolydian)
0.90
F major conf
122
median BPM
94
tempo shifts
0.742
window score

The contrast with Tweezer is the story. Where Tweezer sits on one bluesy C floor and builds through rhythm, Golden Age opens with a flat-seventh Mixolydian lean and then commits to clean F major, the band choosing resolution over tension. The key detector is very sure here at 0.90, which fits a jam that picks a bright tonal home and stays in it while the tempo pushes near 122.

Window timeReadingConfRunner-up
0:00-0:52 F Mixolydian0.71 Bb Ionian (0.71)
0:52-2:02 F Ionian0.68 G Dorian (0.61)
2:02-3:13 F Ionian0.68 G Dorian (0.61)
3:13-4:24 F Ionian0.68 G Dorian (0.61)

Run ranking

Best 4.5-min window in each complete-stem jam. The queue, in priority order.

#JamLengthBest windowScoreDriver
104-23-tweezer 15:307:45–12:15 0.772energy
204-18-golden-age 11:556:30–11:00 0.742energy
304-27-timber 9:104:00–8:30 0.723novelty
404-19-chalk-dust-torture-2 9:254:45–9:15 0.711energy
504-27-twist 7:403:00–7:30 0.669energy

Method

Per-second composite of energy (0.30), harmonic flux (0.30), spectral novelty (0.20), instrument interplay (0.20), z-scored and smoothed. Window score = mean + 0.5Β·std, so it rewards windows that are both active and dynamic. Built on numpy + soundfile + scipy (librosa is broken in this env). Mode and key from the phish-anatomy-of-a-jam skill's mode_timeline.py + jam_analyzer.py, run on a window mix summed straight from the six stems for perfect alignment.