# AIDEA — AI Idea Validation

> For the full version with detailed explanations, see: https://aidea.life/llms-full.txt

> AIDEA is a mobile app that validates your idea with 100 NVIDIA Nemotron personas, delivering 8 metrics — understanding, need fit, curiosity, trial intent, repeat intent, pay intent, trust concern, and switching friction — plus risk hypotheses, a validation plan, and raw persona responses.

## What AIDEA Does

AIDEA is an idea validation app for the earliest stage of product thinking. You write your idea in one paragraph, specify a target audience and situation, and AIDEA runs it through AI personas selected from the NVIDIA Nemotron dataset. Each persona responds independently from their own perspective. Results are deduplicated, aggregated, and structured into a detailed validation report.

AIDEA does NOT replace real user interviews or market research. It is a "first mirror" — a tool to find weaknesses and refine hypotheses before talking to real users.

## Quick Validation (v0.2.0)

One sentence is all it takes. Instead of completing four separate steps (idea type → target audience → pain points → detailed description), users write a single sentence. AI automatically estimates the idea type, target audience, region, and age group.

Key capabilities:
- **AI auto-estimation** — Detects idea type, target, region, and age group from one sentence
- **Clarity meter** — Real-time indicator showing whether the input is ready for validation
- **AI follow-up questions** — Targeted questions that sharpen the idea before validation begins
- **A·I·D·E·A step labels** — Visual labels guiding users through each input stage

## 8 Scored Metrics

Each metric is scored independently and shown separately (not collapsed into a single score):

1. **Understanding** (이해도) — How quickly people grasp what the idea is once explained
2. **Need fit** (필요 부합) — Whether the idea feels like it addresses a real, existing need
3. **Curiosity** (호기심) — How strongly people want to learn more about it
4. **Trial intent** (시도 의향) — Willingness to actually try it out
5. **Repeat intent** (재사용 의향) — Likelihood of coming back after first use
6. **Pay intent** (지불 의향) — Willingness to pay for the value offered
7. **Trust concern** (신뢰 우려) — Level of concern about whether it can be trusted
8. **Switching friction** (전환 장벽) — Perceived barrier to switching from existing alternatives

## Additional Report Sections

Beyond the 8 scored metrics, each report includes:

- **Risk hypotheses** — Where the idea could fail, structured as testable hypotheses
- **Validation plan** — Concrete next steps for what to check with real users
- **Persona voices** — Raw responses grouped by sentiment, so you can read what individual personas actually said

## Bias Mitigation Pipeline

AIDEA applies multiple layers of statistical bias mitigation to prevent groupthink and ensure diverse perspectives:

1. **Three-tier stratified reservoir sampling** — Personas are selected using strict, loose, and minimum diversity thresholds across demographic and psychographic dimensions
2. **Seeded Fisher-Yates shuffle** — Deterministic PRNG ensures reproducible persona ordering while maintaining randomness
3. **Response deduplication** — Duplicate or near-identical responses are detected and collapsed by frequency analysis
4. **Weighted score aggregation** — Final metric scores account for response quality and diversity weighting
5. **Prompt-level diversity enforcement** — The survey prompt itself includes instructions that enforce independent thinking and prevent anchoring bias

## Technical Details

- **Persona source**: NVIDIA Nemotron-Personas dataset (CC BY 4.0 license by NVIDIA)
- **LLM engine**: Google Gemini 2.5 Flash Lite (for persona response generation)
- **Panel size**: 30 or 100 personas per survey (expanding to 500)
- **Platform**: iOS (Android coming soon)
- **Pricing**: Free to start
- **Backend**: Firebase Cloud Functions (Node.js + TypeScript)
- **Data storage**: Firebase Firestore

## Key Facts

- AIDEA is designed for solo founders, product managers, and designers who need to quickly validate an idea direction before committing to real user research
- The app is bilingual: Korean (default) and English
- All persona data is sourced from the NVIDIA Nemotron-Personas dataset under CC BY 4.0 license
- The validation report is structured to help users identify what to check next with real interviews, not to replace those interviews

## Links

- Website: https://aidea.life
- App Store: https://apps.apple.com/app/aidea/id6767673933
- Privacy Policy: https://aidea.life/privacy
- Terms of Use: https://aidea.life/terms
- Contact: eifercompany@zohomail.com
- Developer: EIFER