# Logo method

A logo's job is to be the brand's most distilled, distinctive signature — and to still work at
16px, in one color, on light and dark. Get there with a concept first, then disciplined
execution.

## 1. Find a motif (concept before form)

Don't start by drawing letters. Start by mining meaning:

- The **name** — abbreviations, initials, a syllable, a hidden word.
- **What the business does** — a verb, a tool, a gesture, an object.
- **Double meanings** — the strongest marks say two things at once.

Worked example: **AlphaBlue** → the code tag `<AB/>`. It reads as *code* (angle brackets +
self-closing slash) **and** as *A/B testing* (the two initials). One shape, two meanings, and
it's ownable. Aim for a concept the user can retell in a sentence. A concept beats a
nice-but-arbitrary monogram every time.

## 2. Generate distinct concepts (3-4), on light AND dark

Produce several genuinely different constructions — not four tweaks of one idea. Vary:

- **Literal ↔ abstract**: a recognizable object vs. a pure geometric mark.
- **What carries it**: a symbol mark, a wordmark, an enclosed tile/badge.
- **Bracketing/框景**: freestanding vs. contained.

Render **every** concept on **both a light and a dark background** — a mark that only works on
one is not finished. Label each with a one-line description of its idea and best use.

Recommend a direction, but let the user pick. Offer a combination when it's natural (e.g. one
concept as the horizontal wordmark + a compact one as the app-icon/favicon).

## 3. Refine — and fix the specific problem

Once a direction is chosen, iterate on *its* weakness. The failure modes you'll hit most:

### The cheap two-color clash
A mark using **two colors from opposite temperatures (cold + warm) at equal weight and full
saturation** almost always looks cheap and loud. The three fixes (use one or combine):

1. **Mute + match value** — deepen/desaturate both so they share one lightness. (Bright blue +
   neon coral → cobalt + terracotta.) Keeps the two-color idea but kills the clash.
2. **Demote one color** — make the mark mostly one color and let the second appear as a single
   small accent (one stroke, one dot, the slash).
3. **Single-accent restraint** — the whole mark in ink, exactly one element in the accent. The
   most premium and most reliable.

### Over-literal / weak concept
If it's just "the name in a color," it's decorative, not conceptual. See the evaluation notes.

### Illegible small / one-color
Test the mark at ~16px and in pure `currentColor`. If detail vanishes, simplify.

## 4. Evaluating when the user is torn

When the user can't decide (or scores an option "70/100 but can't say why"), **don't just
generate more options** — give an objective, principled read. The most useful principles:

- **A logo need not literally depict its name.** Matching the mark's color to the name
  ("...Blue" → make it blue) is a weak, decorative idea — it repeats known information and adds
  nothing. Apple's logo isn't apple-red. The name already says the word.
- **Distinctiveness comes from contrast with the surrounding UI, not from matching it.** If the
  whole product UI is blue, a blue mark *disappears into it*. A warm mark in a blue UI becomes
  the signature precisely because it doesn't match. Differentiate, don't echo.
- **Restraint + one deliberate spark is the premium formula.** Large calm area + one precise
  accent = a heartbeat. Filling that role with a "safe" expected color removes the heartbeat.
- **The mark should carry the brand's *difference*, not its category's *convention*.** A techy
  business doesn't need a techy-blue mark to prove it's techy — that's the category cliché. The
  mark can instead say what makes it different (warmth, craft, humanity).

### Resolving name-vs-mark tension
If the literal name-color genuinely matters to the user, you don't have to choose: keep the
**symbol mark** restrained/distinctive, and let the name-color live only in the **wordmark
lockup** (e.g. the word "Blue" set in blue, sitting next to a non-blue mark). The mark stays
premium; the name's meaning still lands in the lockup. Best of both.

## 5. Before production

Confirm: the chosen construction, the exact hex for every part, letter spacing/weight, and the
divider/spacing of the lockup. Then move to `production.md` to outline and export.