✅ Abordagem A: Paleta concreta com hex (AEFRM)
Fornecer hex codes explícitos elimina ambiguidade. O exemplo AEFRM (silver-gray frio) funciona para fintech, enterprise, SaaS sério.
Prompt canônico — Abordagem A
Build a landing page for [AEFRM - a quantitative trading
platform for institutional clients].
Use this exact color palette:
- Base (lightest): #E9ECEC
- Surface: #C9D2D4
- Midtone: #8C9A9E
- Accent / text: #44545B
- Deep (darkest): #11171B
Do NOT introduce any warm tones, creams, or earth colors.
Do NOT use serif fonts for body text.
Use IBM Plex Sans for UI text and IBM Plex Mono for numbers.
Typographic hierarchy:
- H1: Plex Sans 600, tight tracking, 64px
- Body: Plex Sans 400, 16px, 1.5 line-height
- Numbers: Plex Mono 500 (financial figures always monospace)
Restraint:
- Max 2 colors per section
- No gradients
- No rounded corners greater than 4px
- No drop shadows✅ Abordagem B: 4 direções antes de construir
Forçar divergência antes da convergência. Evita que o modelo pule direto para o default.
Prompt canônico — Abordagem B
Before building anything, propose 4 distinct visual directions
for this product. Each direction should have a clearly different
mood, palette, and typographic character — not 4 variants of the
same idea.
For each direction, specify:
1. Mood in one sentence (e.g., "technical precision" vs "soft
editorial" vs "playful high-contrast" vs "monochrome severity")
2. Full color palette with hex codes (5 colors)
3. Display font + body font (Google Fonts names, no serifs by
default unless justified)
4. One signature design choice that makes this direction unique
(e.g., "grid lines visible everywhere", "oversized numerals",
"monochrome photography")
5. One reference product or site this direction evokes
At least one direction must deliberately AVOID cream/beige/
terracotta/warm tones. At least one must avoid serif headings.
At least one must be high-contrast / editorial / bold.
After presenting all 4, STOP. Wait for me to pick one before
you write any code. Do not proceed without confirmation.Por que funciona: "at least one must avoid cream" força o modelo a produzir alternativa real, não um "cream light" disfarçado.
✅ Abordagem C: Snippet anti-slop <frontend_aesthetics>
A abordagem mais robusta das três — cola em claude.md de projeto frontend e rodeia todos prompts.
<frontend_aesthetics>
Do not produce generic AI-generated aesthetic output. Avoid the
following defaults unless the product domain explicitly asks for
them (editorial, hospitality, wellness, artisanal food):
- Background cream / beige / off-white with warm undertone
(e.g. #F4F1EA, #F8F5EC, #FAF6EB)
- Serif editorial display fonts as the primary heading
(Fraunces, Playfair Display, Recoleta, DM Serif)
- Terracotta / burnt-orange / amber accent colors
(e.g. #C9926F, #B8814E, warm oranges)
- Italicized serif headlines on soft backgrounds
- Sparse layouts with generous whitespace and centered content
- Decorative ornaments (swashes, leaves, hand-drawn dividers)
Instead, before you write any code, make concrete, intentional
choices:
1. Name the product domain (fintech, SaaS, health, dev tools, etc.)
2. Pick one mood adjective (precise, playful, severe, friendly)
3. Propose a palette with 5 explicit hex codes AND justify why
each belongs to that mood
4. Propose a primary font and justify (reference a brand that
uses it well)
5. Name at least 2 visual choices that deliberately break the
AI default (e.g., "all-lowercase navigation", "monospace
pricing", "no rounded corners", "aggressive grid")
If your first draft matches the defaults above, stop and
regenerate. Do not ship the default.
</frontend_aesthetics>🎨 Construindo paletas frias (silver/blue/black)
Recipe para paletas frias que comunicam precisão. 5 tons do claro ao escuro, opcionalmente 1 accent cromático.
Template
Regra de saturação: em fintech/healthcare, saturação alta = alarmante. Use accent apenas em 1–2% da tela. Resto é neutro.
🔤 Tipografias fora do óbvio
Inter virou cliché. Fraunces é o default que você quer quebrar. Lista curada:
| Fonte | Para | Mood |
|---|---|---|
| IBM Plex Sans | SaaS B2B, fintech | Engenharia, honesta |
| Söhne | Editorial moderno, enterprise | Suíça contemporânea |
| Neue Haas Grotesk | Dashboard, enterprise | Clássica, neutra |
| JetBrains Mono | Dev tools, code-heavy | Técnica, precisa |
| GT America | Startup confiante | Americana geométrica |
| Berkeley Mono | Terminal UIs, data platforms | Nostalgia serifada-mono |
⚠️ Especifique no prompt. Se você só pedir "nice font", o 4.7 volta para Inter ou Fraunces.
🎬 Micro-interações como diferencial
O que separa uma landing AI-generated de uma feita por time sério: motion design pensado.
// Especifique explicitamente no prompt:
"Interactions:
- Primary button hover: scale(1.02) + background shift from
#0066FF → #0052CC over 120ms ease-out.
- Card hover: translateY(-2px) + shadow shift. Not scale.
- Numbers in hero: count-up animation on scroll into view,
1200ms, easeOutExpo.
- Navbar: transforms from transparent to solid on scroll Y > 64
with 160ms backdrop-blur ramp.
Easing library: Framer Motion with custom cubic-bezier(0.22, 1, 0.36, 1).
Duration budget: nothing over 400ms. Sub-100ms for micro feedback."🖼️ Estudo de caso: dashboard fintech em paleta fria
Aplicação conjunta das 3 abordagens. AEFRM paleta + IBM Plex + micro-interações.
📋 Resumo
Próximo Módulo:
4.7 — Projeto Final Integrador