Sourdough Starter Whisperer
Overview
A sourdough starter is a livestock of yeast + lactic-acid bacteria, and it dies or underperforms for boring, predictable reasons: fed too cold, fed too late, wrong ratios, or misread signals. Bakers lose weeks to "my starter never doubles" when the fix is a 26°C spot and a 1:5:5 feed for two days.
This skill is the starter's instrument panel:
- Feeding math — ratios that scale (1:5:5, 1:2:2), retention math (keep X g, discard down), and schedules for counter (24°C, 26°C, 28°C) vs fridge life.
- Hydration calculator — convert between 100% hydration and stiff (50–65%) or liquid (120–166%) levains; compute flour/water for any target from what you have.
- Readiness — expected peak time by temperature and ratio; the float test's false positives; rise-tracking in a marked jar.
- Diagnosis engine — 10 symptom clusters (no rise, acetone/nail- polish smell, gray hooch layer, mold vs kahm, sluggish after fridge, gummy collapse) each with a cause and a numbered recovery protocol.
- Vacation mode — dry-chip and fridge-hibernation protocols with revival timelines.
- Levain build planner — from starter to peak-aligned levain for a specific bake time, off a 1:1:5 build at your kitchen temperature.
When to Use
- Keeping any sourdough starter (wheat or rye) alive and vigorous
- Starter looks dead / smells like nail polish / has liquid on top / rose and collapsed / never doubles
- Planning travel: how to park a starter for 1–8 weeks
- Building a levain timed to tomorrow's bake
- Converting starter hydration for a recipe (stiff vs 100% vs liquid)
Don't use for: commercial-yeast bread math (no starter involved), gluten-free starter troubleshooting (different microbial stable — rye adaptation advice only), or kneading/shaping technique.
Quick Start
python3 scripts/starter.py feed --temp 24
python3 scripts/starter.py feed --temp 28 --style stiff
python3 scripts/starter.py hydration --have 100 --target 65 --mass 200
python3 scripts/starter.py readiness --temp 24 --ratio 5
python3 scripts/starter.py diagnose --symptom "smells like acetone"
python3 scripts/starter.py vacation --weeks 3
python3 scripts/starter.py levain --bake 07:30 --temp 22
Stdlib-only, no network.
Core Numbers (what the math encodes)
| Parameter | Value | Meaning |
|---|---|---|
| Standard feed | 1 : 5 : 5 (starter:flour:water by weight) | peak in ~10-14 h at 24°C |
| Equal feed | 1 : 1 : 1 | fast peak (4-6 h at 24-26°C), acid accumulation — good for revival, bad as routine |
| 100% hydration | equal flour and water | the "default" liquid starter most recipes assume |
| Stiff levain | 50-65% hydration | slower, more acetic (tangier), easier timing |
| Peak at 24°C | ~2x rise, 10-12 h after 1:5:5 | jar marking makes this visible |
| Peak at 28°C | ~2x rise, 6-8 h | warm kitchens mature starters fast; feed 2x/day |
| Fridge storage | feed 1:2:2, wait 2 h, refrigerate | keeps ~2 weeks before a revival cycle |
| Hooch | dark liquid on top | underfed — pour off, feed; normal after fridge |
| Acetone smell | yeast stressed/starved | 2-3 days of 1:5:5 at 24-26°C fixes it |
| Float test | floats in water ≈ ready | FALSE PASS right after feeding; trust rise marks |
Workflow
- Baseline your kitchen:
feed --temp T(T = where the jar actually lives, not the room's daytime high) → your feeding rhythm + ratios. - Mark the jar at feed time with a rubber band; target 2x by next
feed.
readiness --temp --ratiogives expected peak windows. - Diagnose before adjusting: any off-signal →
diagnose --symptomfirst; most "dead" starters need 2-3 good days, not funeral rites. - For tangier/stronger bread use
hydration --target 60to build a stiff levain; for milder and faster, stay at 100% or go 120%. - Travel:
vacation --weeks N→ fridge or dry-chip protocol with a revival timeline and a first-bake-ready date. - Bake day:
levain --bake HH:MM --temp T→ the 1:1:5 build start time so the levain peaks exactly when mixing starts.
Common Pitfalls
- Feeding by volume (cups) instead of weight. Hydration and ratios only exist by mass; a spoon of flour and a spoon of water are ~50% off from 100% hydration.
- Equal 1:1:1 feeds forever. Peaks fast, acidifies fast, then the "my bread stopped rising" mystery. Routine feeding is 1:5:5 (or 1:10:10 for a very warm spot).
- Cold kitchen, warm-kitchen expectations. At 20°C a 1:5:5 peak can take 16-20 h; bakers read it as "dead". Move the jar (top of fridge, lamp-lit cupboard) before panicking.
- Reading hooch as disaster. Dark liquid = underfed, not dead. Pour off, feed normally, 24 h later it's fine. Only pink/orange fuzz or fuzzy colored spots = mold = start over (or salvage from below).
- Trusting the float test blindly. Freshly fed starters float from trapped air bubbles; a starter 3 h past peak can also float. The jar mark + expected peak window is the real instrument.
- Refrigerating right after feeding. The culture drops dormant before consuming fresh flour → sour, weak revival. Feed 1:2:2, let it work 2 hours at room temp, then refrigerate.
- Revival expectations. A fridge-hibernated starter needs 2-3 feed cycles (1-2 days) before it doubles reliably; baking on day-1 revival gives dense loaves.
- Rye and wheat swaps without recalibration. Rye ferments faster (more enzymes/minerals) — feed a rye starter on a shorter cycle or cut ratios to 1:7:7 equivalent timing.
Verification Checklist
- Feeding by WEIGHT, ratios from the tool's plan for your temperature
- Jar marked at feed; 2x rise confirmed against expected peak window
- Any anomaly run through
diagnosebefore changing routine - Refrigeration follows the feed→2h→chill sequence
- Revivals given 2-3 cycles before judgment
- Levain build timed to bake time, confirmed by rise mark, not clock alone
Safety Notes
- Mold (fuzzy, colored, on top) = discard; pulling starter from under a moldy layer sometimes works but is not worth it with flour this cheap.
- Everything else — hooch, acetone smell, gray tint — is recoverable.
- Pink/orange streaks = unwanted bacteria have won; start fresh.