Communication Gym (Product One)
Designed and prototyped an AI-assisted communication coaching application called Communication Gym (Product One). Built a functional no-code prototype on Bubble.io with a 30-Day Topic Ladder, 5 AI evaluation modes, and r
Executive Summary
Designed and prototyped an AI-assisted communication coaching application called Communication Gym (Product One). Built a functional no-code prototype on Bubble.io with a 30-Day Topic Ladder, 5 AI evaluation modes, and raw-attempt-based performance scoring system, all under a zero-budget constraint.
2The Problem
Professional communication is a critical career skill, yet most solutions are passive (courses, videos) or unstructured (generic AI chatbots). Professionals struggle with clarity, confidence, repetitive language, and weak sentence flow without a structured practice mechanism that provides honest, unsentimental feedback.
Design a product that forces active communication production rather than passive learning. Provide AI-powered critique that is honest and unsentimental. Separate user performance from AI assistance to maintain measurement integrity. Build everything under a zero-budget constraint using free-tier tools.
Operating Constraints
Zero budget all tools had to be free-tier. No development team solo product designer/builder. AI integration limited to free-tier APIs. Product scope had to match available free-tier limitations while still demonstrating the core concept.
3Objective
- •Design a product that forces active communication production rather than passive learning. Provide AI-powered critique that is honest and unsentimental. Separate user performance from AI assistance to maintain measurement integrity. Build everything under a zero-budget constraint using free-tier tools.
4My Role & Responsibility
I Owned
- •Designed a closed practice-and-feedback loop with 7 mechanisms. Created a 30-Day Topic Ladder progressing from self-introduction through leadership and persuasion. Developed 5 AI Judge Modes including a Harsh/Olympic Judge for unsentimental critique. Implemented the Muscle Score concept: scoring based strictly on raw human attempts, with AI corrections never inflating the score. Built working Bubble.io prototype with 4-table database schema.
I Contributed To
- •Explored Hugging Face, GLM/Zhipu AI, and Mistral API integrations for AI evaluation modes
Team Delivered
- •Designed a closed practice-and-feedback loop with 7 mechanisms. Created a 30-Day Topic Ladder progressing from self-introduction through leadership and persuasion. Developed 5 AI Judge Modes including a Harsh/Olympic Judge for unsentimental critique. Implemented the Muscle Score concept: scoring based strictly on raw human attempts, with AI corrections never inflating the score. Built working Bubble.io prototype with 4-table database schema.
- •Designed a closed practice-and-feedback loop with 7 mechanisms. Created a 30-Day Topic Ladder progressing from self-introduction through leadership and persuasion. Developed 5 AI Judge Modes including a Harsh/Olympic Judge for unsentimental critique. Implemented the Muscle Score concept: scoring based strictly on raw human attempts, with AI corrections never inflating the score. Built working Bubble.io prototype with 4-table database schema.
- •Explored Hugging Face, GLM/Zhipu AI, and Mistral API integrations for AI evaluation modes
- •Designed a closed practice-and-feedback loop with 7 mechanisms. Created a 30-Day Topic Ladder progressing from self-introduction through leadership and persuasion. Developed 5 AI Judge Modes including a Harsh/Olympic Judge for unsentimental critique. Implemented the Muscle Score concept: scoring based strictly on raw human attempts, with AI corrections never inflating the score. Built working Bubble.io prototype with 4-table database schema.
5Approach & Method
Designed a closed practice-and-feedback loop with 7 mechanisms. Created a 30-Day Topic Ladder progressing from self-introduction through leadership and persuasion. Developed 5 AI Judge Modes including a Harsh/Olympic Judge for unsentimental critique. Implemented the Muscle Score concept: scoring based strictly on raw human attempts, with AI corrections never inflating the score. Built working Bubble.io prototype with 4-table database schema.
Why This Approach
The separation of AI assistance from user performance scoring (Muscle Score) was the single most important design decision. Without this separation, the product loses measurement integrity users cannot distinguish what they actually produced from what the AI improved. Chose Bubble.io over custom development because the zero-budget constraint made no-code the only viable path to a functional prototype.
6Evidence & Artifacts
7Outcome & Results
- •Completed product concept and architecture. Designed 30-Day Topic Ladder and Guideline Card system. Prototyped 5 AI evaluation modes. Implemented functional Bubble.io prototype with User, DailyAttempt, DayContent, and PracticeSession database tables. Established Raw Attempt/Muscle Score scoring logic. Explored AI API integrations. Not commercially launched due to resource constraints.
- •The Muscle Score concept separating raw human performance from AI assistance represents a novel approach to measurement integrity in AI-assisted learning products.
Key Results
- •Completed product concept and architecture. Designed 30-Day Topic Ladder and Guideline Card system. Prototyped 5 AI evaluation modes. Implemented functional Bubble.io prototype with User, DailyAttempt, DayContent, and PracticeSession database tables. Established Raw Attempt/Muscle Score scoring logic. Explored AI API integrations. Not commercially launched due to resource constraints.
Impact
- •The Muscle Score concept separating raw human performance from AI assistance represents a novel approach to measurement integrity in AI-assisted learning products.
8What I Learned
- •The separation of AI assistance from user performance scoring (Muscle Score) was the most critical design decision in the entire product. Product scope must match available resources; the gap between vision and free-tier limitations prevented production readiness but the core concept was successfully demonstrated. AI should serve the product methodology, not become the product identity.