Translating Ideas to Modality

Quick Start

Given an idea, first classify its target modality, then produce the artifact directly — no meta-commentary, just the translated output plus a one-line rationale.

Example request: "I have an idea for a user onboarding flow with 3 steps and a drop-off check."

Output:

MERMAID
flowchart TD
    A[User Signs Up] --> B[Step 1: Verify Email]
    B --> C{Verified?}
    C -- No --> D[Send Reminder] --> B
    C -- Yes --> E[Step 2: Complete Profile]
    E --> F[Step 3: First Action Prompt]
    F --> G{Action Taken?}
    G -- No --> H[Drop-off Tracked]
    G -- Yes --> I[Onboarded]

Rationale: Branching logic with decision points is clearer as a flowchart than as prose.

Workflow

Progress:

• Step 1: Identify the core structure of the idea (sequential, relational, visual, quantitative, behavioral/logical)
• Step 2: Match structure to modality (see mapping table below)
• Step 3: Strip the idea to its essential components — actors, states, relationships, constraints
• Step 4: Translate into target format using that modality's native conventions
• Step 5: Validate the translation preserves intent (no information loss, no invented details)
• Step 6: Deliver artifact + 1-line rationale for modality choice

Modality mapping:

Idea structure	Target modality	Why
Steps, branches, state transitions	Flowchart / diagram (mermaid, graphviz)	Visualizes control flow
Discrete logic, transformations, algorithms	Code (pseudocode or real language)	Executable precision
Multiple attributes across multiple items	Table	Compact comparison
Visual scene, mood, style, composition	Image-gen prompt	Needs visual vocabulary, not logic
Hierarchies, nested categories	Tree diagram / nested list	Shows containment
Numeric relationships, trends	Chart spec (e.g. matplotlib/vega-lite config)	Quantitative clarity

Examples

Example 1: Input: "A character — a weary cyberpunk detective standing in neon rain, cinematic lighting." Output:

Prompt: A weary cyberpunk detective in a long trench coat, standing in heavy neon-lit rain at night, reflections on wet pavement, cinematic lighting, shallow depth of field, moody teal and magenta color grade, ultra-detailed, 35mm film still
Negative prompt: cartoon, flat lighting, low detail, daylight

Rationale: Visual/atmospheric idea maps to an image-gen prompt, not code or diagram.

Example 2: Input: "I want to compare three pricing plans by features and cost." Output:

Feature	Basic	Pro	Enterprise
Price/mo	$9	$29	$99
Users	1	10	Unlimited
Support	Email	Priority	Dedicated
API Access	No	Yes	Yes

Rationale: Multi-attribute comparison across items is clearest as a table.

Example 3: Input: "Idea: validate that a password has 8+ chars, one digit, one uppercase letter." Output:

Python
import re

def is_valid_password(pw: str) -> bool:
    return bool(re.match(r'^(?=.*[A-Z])(?=.*\d).{8,}$', pw))

Rationale: Discrete validation logic translates directly to executable code.

Best Practices

• Choose modality based on the idea's structure, not the user's wording — a "diagram" request that's actually tabular data should still become a table if that's clearer.
• Keep translations lossless: every constraint or detail in the source idea must appear in the output.
• For image prompts, always separate subject, style, lighting, and composition into distinct descriptive clauses.
• For code, prefer the simplest language/pseudocode that proves the concept unless a specific stack is requested.
• When an idea spans multiple structures (e.g., a process with quantitative output), produce multiple artifacts rather than forcing one modality to do everything.

Common Pitfalls

• Don't add invented details (extra steps, colors, data) not implied by the source idea.
• Don't default to prose explanation when a structured modality was clearly better suited — translate, don't describe.
• Don't produce an image prompt that's just a literal restatement of the idea without visual vocabulary (style, lighting, composition, medium).
• Don't use a table for sequential/branching processes — that's a flowchart's job.
• Don't skip the rationale line; it signals why that modality was chosen, which matters when the mapping isn't obvious.