Mochi

What the company appears to be: Mochi is a self-reported iPhone and Apple Watch app designed for people who are hard of hearing. The author states it provides real-time captions, speaker-aware cues, AI-powered catch-up summaries, and evidence-linked recaps. It uses OpenAI Realtime, WhisperKit, FluidAudio, and GPT-5.6.

What changed: This is a single-person hackathon project submitted to the OpenAI 2026 hackathon. The author describes building an end-to-end native app with audio pipelines, AI integration, and watchOS support in a short timeframe.

The single most important open question: Is there any evidence of user testing or real-world adoption beyond the author's own experience? The description states no revenue, customers, or traction data exist beyond the project submission.

The description states Mochi is an iPhone and Apple Watch app for people who are hard of hearing. It combines:

• Multilingual live captions (OpenAI Realtime with WhisperKit fallback)
• Speaker-aware cues and name mentions
• AI-powered catch-up summaries (GPT-5.6)
• Conversation recaps with searchable transcripts
• Evidence-linked recaps that connect AI output to real transcript segments
• Voice Lift (optional headphone assistance)

The app is built in SwiftUI for iPhone and Apple Watch, using Codex and GPT-5.6 for development.

The author states Mochi helps people "stay in the conversation" with real-time captions, cues, and AI catch-up. The positioning emphasizes:

• Accessibility for hard-of-hearing users
• Privacy (local processing where possible)
• Accuracy and evidence-based summaries
• Companionable experience ("calm and companionable")

The claim evolution appears to be from a hackathon prototype to a working end-to-end app with multiple features, but no evidence of market positioning or customer feedback beyond the author's own experience.

The description states Mochi is for "people who are hard of hearing" and specifically mentions the author's personal experience with unreliable hearing. The target user appears to be someone who:

• Is physically present in conversations
• Needs to track name mentions, questions, and important details
• Values privacy and accuracy
• Uses iPhone and Apple Watch

No evidence of customer segmentation or specific personas beyond "hard-of-hearing users."

Not evidenced. The description does not mention any pricing model, monetization strategy, or business model.

The description states:

• Native SwiftUI app for iPhone and Apple Watch
• Audio pipeline with local recording and streaming captions
• OpenAI Realtime for low-latency captions with WhisperKit fallback
• FluidAudio Sortformer for on-device speaker diarization
• GPT-5.6 for semantic layer (catch-up, recaps)
• Node.js server for API key management and credential minting
• On-device processing where possible
• Evidence linking AI output to real transcript segments

The author notes challenges with latency vs accuracy trade-offs and audio pipeline coordination.

Not evidenced. The description states this is a hackathon project submitted to the OpenAI 2026 hackathon, with no revenue, customers, or traction data beyond the project submission itself.

Not evidenced. No mention of existing competitors or market analysis in the description.

• Single-person development team (1 member)
• No evidence of user testing beyond author's own experience
• No revenue, customer, or traction data
• GPT-5.6 is claimed but not verified as actual implementation
• Self-reported technical claims without independent verification
• No evidence of market validation or product-market fit

1. What specific user testing has been conducted beyond the author's own experience?
2. How does the app handle edge cases in real-world noisy environments?
3. What is the actual latency and battery impact in real usage?
4. Has the team considered regulatory or medical device compliance issues?
5. What are the plans for scaling beyond a single developer?
6. How will user privacy be maintained at scale?