Coffee Brew Lab

Overview

A dial-in instrument for manual coffee brewing. Two engines: (1) recipe math — SCA golden ratio (1:15–1:18), method-specific grind/temperature/dwell profiles, bloom and pulse schedules, temperature-corrected for altitude; (2) taste diagnosis — a defect→cause tree that converts "bitter, hollow, sour" into concrete parameter moves, plus a brew log that tracks every shot and identifies which variable actually correlates with your best cups.

When to Use

• "How much coffee/water for N cups?" — ratio math for any method
• "My coffee tastes bitter/sour/weak/harsh" — diagnosis tree
• Dialing in a new bag, grinder, or method (pour-over, AeroPress, French press, cold brew, moka, espresso)
• "What grind size / temperature for X method?" — method profiles
• Keeping a brew log and finding personal sweet spots

Don't use for: where to buy beans, milk-frothing art, or cafe business ops.

Method Profiles (summary; full table in references)

The Diagnosis Tree (core of the skill)

Bitter / harsh / drying aftertaste = OVER-extraction → coarser grind OR lower temperature OR shorter dwell OR less agitation (one variable at a time).

Sour / grassy / salty / thin-sour = UNDER-extraction → finer grind OR hotter water OR longer dwell OR more agitation.

Weak / watery but not sour = ratio problem → more coffee (lower ratio number, e.g. 1:16 → 1:15), not a grind change.

Both bitter AND sour = channeling or uneven grind → fix distribution, pulse pouring, check grinder for bimodal distribution; do NOT chase with temperature.

Hollow / boring / papery = stale beans or under-dose → check roast date (ideally 7–45 days), then dose.

Rule: change ONE variable per cup, log it, keep the winner.

Numbered Workflow

1. recipe --method v60 --cups 2 --ratio 16 → grams coffee, water, bloom, pour schedule, grind, temperature (altitude-corrected if --altitude).
2. Brew it. Taste against the descriptors above.
3. log add the brew: method, ratio, grind setting, temp, dwell, and a 1–10 score plus defect tags (bitter/sour/weak/balanced).
4. diagnose bitter (etc.) → get the prescribed single-variable move.
5. Apply the move, brew, log again. 2–4 iterations converge for most setups.
6. log best → your top cups and the parameters they share; log analyze → simple regression of score vs each numeric variable.

Commands

python3 scripts/brew_lab.py recipe --method v60 --cups 2
python3 scripts/brew_lab.py recipe --method french-press --grams 30
python3 scripts/brew_lab.py recipe --method cold-brew --liters 1
python3 scripts/brew_lab.py diagnose bitter
python3 scripts/brew_lab.py diagnose sour-and-bitter
python3 scripts/brew_lab.py log add --method v60 --grams 15 --water 240 \
  --grind 4.5 --temp 93 --dwell 180 --score 7 --tags bitter
python3 scripts/brew_lab.py log best
python3 scripts/brew_lab.py log analyze

Brew log: brew_log.json next to the script (--store to override).

Pitfalls

1. Changing two variables at once — you learn nothing. The log enforces nothing; discipline is yours. Change grind OR temp OR time, never two.
2. Chasing bitterness with dose — bitterness is extraction; dose fixes strength. Wrong axis, wasted beans.
3. Trustiing thermos-kettle numbers — many kettles read 3–6°C high; verify once with a real thermometer and adjust mentally forever.
4. Grinder settings aren't portable — "4.5 clicks" is YOUR grinder's language. Log your own numbers; never copy someone else's.
5. Judging boiling-hot coffee — high temps mute sweetness and amplify bitterness. Taste at ~60–65°C, slurped aerated.

Example

New Ethiopian light roast tastes "sour and thin": diagnosis says under-extraction → grind 2 clicks finer, same everything else → next cup "sour mostly gone, slightly hollow" → +1 click finer and temp 93→95°C is NOT allowed (two variables) → grind only → "balanced, juicy, 8/10". Log shows best cups cluster at grind 3.5, 94°C, 3:00 — that's the bag's sweet spot, recorded for next time you buy it.