Palate taxonomy v0.5: new type names, and how the lab classifies a palate
We renamed the six palate types so the names describe what we measure, added a plain-English meaning to every report, and moved to answer-weighted classification. Here is the full method.
1. What changed
Our first taxonomy (v0.4) gave each palate a playful name. Testers liked them, but the names said more about mood than about what the lab measures. In v0.5 each type is named for the reading that dominates its profile, with a short code you can compare with friends.
| Code | v0.5 name | v0.4 name | Dominant reading |
|---|---|---|---|
| S-1 | Sucrophile | Sugar Voltage | Sweet sensitivity |
| A-2 | Acidophile | Sour Outlier | Sour tolerance |
| B-3 | Amarophile | Dark Matter Palate | Bitter tolerance |
| F-4 | Fractophile | Crunch Architect | Crunch demand |
| V-5 | Viscophile | Velvet Melter | Melt preference |
| P-6 | Polymodal | Chaos Palate | Chaos index |
2. Why the names changed
A type name should tell you what we found. The v0.5 names come from Latin and Greek roots used across the sciences: -phile means “lover of,” so a Sucrophile is literally a lover of sugar. The codes are short enough to say out loud (“I’m an A-2”) and stay the same if a name ever changes again.
3. The six types, explained
S-1 Sucrophile
From sucrose (table sugar) + -phile, “lover of”.
Sweet is your dominant reading: your palate responds early and strongly to sugar. Your receptors light up early and stay lit. You want sweetness with a charge behind it.
A-2 Acidophile
From acid + -phile, “lover of”.
Sour is your dominant reading: you have a high tolerance for citric and malic acid, and you like it. You treat citric acid as a love language. Pucker is a feature, not a bug.
B-3 Amarophile
From Latin amarus, “bitter” + -phile.
Bitter is your dominant reading: cacao, coffee and burnt sugar read as depth, not a flaw. Bitter doesn’t scare you. Cacao, burnt sugar, espresso: you read them as depth.
F-4 Fractophile
From Latin fractus, “broken” + -phile.
Crunch is your dominant reading: texture and the snap of the first bite matter most. Texture is the whole point. If it doesn’t snap, crackle or shatter, it didn’t happen.
V-5 Viscophile
From viscous (thick, slow-flowing) + -phile.
Melt is your dominant reading: you want slow, smooth candy like caramel, nougat and truffles. You want slow collapse: caramel, nougat, anything that gives up on your tongue.
P-6 Polymodal
From poly-, “many” + modal, “modes”.
No single reading dominates: you want several taste modes at once, sweet, sour, salty and spicy together. Sweet, sour, salty, fizzy, all at once. You are the reason flavor pairing papers exist.
4. How a palate is classified
The scan. Your phone or laptop analyzes a photo of your tongue on the device and reduces it to a handful of image summaries. From these, the model produces six readings from 0 to 100. We are candid about what this is: a designed experience inspired by real taste research, not a measurement of your taste receptors. Readings are stable for each person and vary by a few points between scans, the way a real instrument would. Nothing about the scan is medical or diagnostic.
Calibration. After the scan you answer six questions: textures, chocolate or fruity, sour tolerance, sugar, what you avoid, and budget. These are real signal. Your answers adjust the readings (for example, your stated sour tolerance pulls the sour reading toward it), which is why your report reflects how you actually eat.
Your type. Your type is the reading that dominates after calibration. If calibration moves another reading close to the top, your type stays the same, so the label you shared doesn’t change under you.
Taster class and papillae index. These are playful labels inspired by research on supertasters and fungiform papillae (Bartoshuk et al., 1994; Hayes et al., 2008). The real relationship between papillae and taste is contested (Fischer et al., 2013).
Compatibility and verdict. Your compatibility score compares your palate with Preview Model #01. About 4 in 5 palates qualify. Your verdict is fixed for your profile, so rescanning or changing answers won’t flip it.
Candy matching. Once you’ve answered at least four questions, your answers outweigh the scan readings in matching. We filter out candies built around allergens or diets you avoid (always check the label), avoid recommending your known favorites, keep one “treat yourself” pick, and learn from your Love, Meh and Nope ratings.
5. What comes next
v1.0 replaces our lab-panel candy ratings with taster data: answers, favorites, ratings and the candies people choose. When enough tasters have contributed, we’ll recompute CandyBench and test whether any image features predict stated preferences. We’ll publish the result either way.
6. Limitations
- The scan is not a scientific instrument, and the type is a fun label, not a diagnosis.
- Calibration relies on what you tell us, and people are not always accurate about their own taste.
- Six types compress a lot of variation. Two Acidophiles can still disagree about Warheads.
Changelog
- v0.5 · Sep 28, 2026 New type names and codes; plain-English meanings on every report; calibration questions between scan and results; answer-weighted candy matching.
- v0.4 · Sep 27, 2026 First public preview: tongue scan, six readings, v0.4 type names, Preview Model #01 verdict.