OneDish

• The description states OneDish is a meal recommendation tool.
• It filters a catalog of 90 demo dishes from ten fictional restaurants.
• It uses user inputs like budget, dietary restrictions, craving, and recent meals to make one decision.
• It shows the elimination process: what was removed, which constraints applied, and why the final dish won.
• The app stores preferences and history locally using IndexedDB through Dexie.js.
• It is a Progressive Web App (PWA) built with React, TypeScript, and Vite.
• It uses OpenAI's Responses API for interpreting natural language into structured fields but does not use AI to choose or rank dishes.
• The backend is optional, built with FastAPI and Pydantic, and can integrate with Foursquare and OpenStreetMap for location data.
• It includes a "Taste Orbit" feature that visualizes meal history as an interactive preference map.
• It has a privacy page showing where each type of data goes, what stays on the device, and what users can delete.

Confidence High — based on detailed self-reported technical and functional description.

• The tagline is: “One decision. The right meal, right now.”
• The author states that most recommendation products respond to choice overload with another list.
• OneDish aims to offer the opposite: one dish with a clear explanation of why it survived.
• It positions itself as a tool that makes decisions transparent and deterministic rather than opaque or AI-driven.
• It emphasizes trust in single recommendations over lists, stating “A single recommendation requires more trust than a list.”
• The author claims that AI is most useful at the boundary between human language and structured software — interpreting ambiguous inputs like “comfort food, but lighter” — while deterministic code enforces strict rules such as allergen exclusions.
• It also positions itself on privacy by integrating explanations directly into the interface.

Confidence Medium — claims are self-reported and not independently verified; no evidence of market positioning or customer feedback.

• The description does not name specific target customers or personas.
• It implies a user base that is tired at 8 p.m., overwhelmed by choice, and looking for clarity in meal decisions.
• It targets individuals who value transparency in how recommendations are made.
• It may appeal to people with dietary restrictions or those seeking consistency in their food choices.
• The app supports offline use and local data storage, suggesting a focus on personal usage rather than enterprise or B2B.

Confidence Low — no evidence of defined customer segments or ICP beyond general user behavior assumptions.

• Not evidenced.
• No mention of pricing models, monetization strategies, or revenue streams in the description.
• The project is described as a demo with fictional restaurants and versioned menu data.
• It does not appear to have any live integrations or commercial partnerships.

Confidence Very low — no business model or pricing information provided.

• Built as a PWA using React, TypeScript, Vite.
• Stores user data locally in IndexedDB via Dexie.js.
• Uses deterministic logic for decision-making and explanations.
• Integrates OpenAI’s Responses API for natural language interpretation only — not for ranking or selecting dishes.
• Optional FastAPI backend with Pydantic validation.
• Can integrate with Foursquare and OpenStreetMap for location data.
• Includes animations to explain the elimination process without pretending to be AI thinking.
• Designed to work offline after first load.
• Supports mobile layout and bilingual content.

Confidence High — detailed technical architecture provided by the author.

• Not evidenced.
• The project is described as a hackathon submission (OpenAI 2026).
• It uses fictional restaurants and versioned menu data.
• No mention of live users, customer feedback, or adoption metrics.
• No evidence of revenue, funding rounds, headcount, or product maturity beyond the demo.

Confidence Very low — no traction or maturity indicators available.

• Not evidenced.
• The description does not reference competitors or market positioning.
• It does not describe how it compares to existing meal recommendation apps or services.
• No mention of competitive advantages or differentiation strategies.

Confidence Very low — no competitive context provided.

• No commercial traction or revenue: The project is a demo with fictional data and no live integrations.
• Single-person team: Only one member (Yanhao Chen) is listed, which raises questions about scalability and execution capacity.
• Limited scope: It currently works only with a fixed catalog of 90 dishes from ten fictional restaurants.
• Unproven AI integration: While it uses OpenAI’s API, it does not appear to use AI for core decision-making or ranking.
• Privacy as UI feature: This is an innovative approach but untested in real-world usage.
• Hackathon project: Not a product in development, but a prototype submitted for competition.

Confidence Medium — risks are inferred from lack of evidence and project nature.

1. What is the plan to transition from demo mode with fictional data to live menu integrations?
2. How do you intend to scale beyond a single developer?
3. Have you tested this with real users or gathered feedback on usability?
4. Are there any plans for monetization or revenue generation?
5. What are your thoughts on privacy compliance and data handling in a production environment?
6. Is there any intention to expand beyond meal recommendations into related lifestyle or health domains?

Confidence Medium — these questions are based on the absence of evidence around traction, scalability, and business model.