KYT(Kiken Yochi Training) Lab — Education That Saves Lives

KYT Lab is described as a web-based educational platform that brings Japanese Kiken Yochi Training (KYT) — a participatory safety culture practice — to global teams through AI-generated workplace scenes and real-time collaborative sessions. The author, a chemist with industrial experience, states the goal is to help vocational educators and workplace safety trainers conduct KYT sessions using digital tools while preserving the method’s core pedagogy: people observe, discuss, and decide together.

The platform uses GPT-5.6 for coaching feedback and gpt-image-2 for generating relevant training scenes, all within a real-time collaborative interface built with React, ElysiaJS, Bun, PostgreSQL, and WebSockets. It supports four rounds of KYT: observe, prioritize, plan, and commit.

The most important open question is whether the described educational approach has any validated traction or adoption beyond this single developer’s prototype. The author claims to have built a working product but does not provide evidence of customers, revenue, usage metrics, or field testing results.

The description states that KYT Lab is a web-based collaborative platform for conducting Kiken Yochi Training (KYT) sessions in teams. It transforms traditional pen-and-paper KYT into an authenticated, real-time digital experience.

Key features include:

• Real-time collaborative session support via WebSockets and Redis pub/sub
• AI-generated workplace scenes using gpt-image-2
• GPT-5.6-powered coaching that provides strengths-first feedback after learners make their own attempt
• Structured four-round KYT process: observe, prioritize, plan, commit
• Evidence markers linked to normalized image coordinates
• Learning records capturing visual evidence, danger statements, votes, decisions, actions, and AI coaching history

The system is described as a monorepo built with React 19, Vite, ElysiaJS on Bun, PostgreSQL, Prisma, TypeScript, Zod schemas, Playwright tests, Docker, Kubernetes, Argo CD, and OpenAI APIs.

The author positions KYT Lab as a tool that brings Japanese safety culture practices to global teams through digital collaboration. The platform is described as:

• Not an AI hazard identifier but a training tool where "AI proposes, people decide"
• Designed for vocational educators, laboratory instructors, workplace safety trainers, supervisors, and team facilitators
• A way to preserve KYT’s pedagogy in digital form without replacing human judgment

The claim evolution shows:

1. Initial inspiration from personal industrial experience
2. Goal to make KYT accessible beyond language and geographic barriers
3. Focus on collaborative learning, not automation or compliance certification
4. Emphasis on attempt-first coaching and human decision-making gates

The description states that KYT Lab is intended for:

• Vocational educators
• Laboratory instructors
• Workplace safety trainers
• Supervisors
• Team facilitators

These users are described as those who want learners to practice hazard recognition together, especially when generic safety materials do not resemble actual work environments.

The ICP appears to be:

• Individuals or organizations responsible for workplace safety training
• Those seeking participatory learning methods that emphasize observation and group discussion
• Users who value structured pedagogy over automated risk assessment

Not evidenced. The description does not contain any information about pricing, monetization strategy, revenue model, or business structure.

The platform is described as:

• Built with React 19, Vite, ElysiaJS on Bun
• Uses PostgreSQL and Prisma for data persistence
• Implements Redis pub/sub and WebSockets for real-time collaboration
• Leverages OpenAI APIs (GPT-5.6, gpt-image-2) with structured outputs
• Includes local or S3-compatible storage, Zod schemas, and Playwright tests
• Deployed using Docker, Kubernetes, Argo CD

The architecture is described as:

Browser → Elysia API → PostgreSQL

├→ Redis pub/sub and WebSockets

├→ Private image storage

└→ OpenAI text and image models

Not evidenced. The description does not contain any information about:

• Customers or users
• Revenue or monetization
• Usage metrics or adoption rates
• Product-market fit or customer feedback
• Production deployment beyond a prototype
• Field testing or validation studies

The author states they have built a working product but provides no evidence of traction.

Not evidenced. The description does not mention:

• Competitors in the safety training space
• Other KYT implementations or digital platforms
• Market size or competitive landscape
• Differentiators from existing tools

Inferences based on self-reported information:

1. Lack of traction: No evidence of users, customers, or revenue; only a prototype built by one person.
2. AI safety boundaries: While the author claims to have implemented responsible AI practices (e.g., attempt-first coaching), these are not independently verified.
3. Single-founder project: The team size is listed as 1, suggesting limited resources for scaling or marketing.
4. Unvalidated educational impact: The description states that this submission demonstrates a working product and workflow but is not a field-efficacy study.
5. High-stakes AI use without third-party validation: Safety training involving AI raises regulatory and ethical concerns, especially if not validated by external experts.

1. What specific feedback have you received from potential users or domain experts in safety education?
2. How do you plan to validate the educational effectiveness of the KYT process in a digital format?
3. Have you conducted any pilot tests with actual teams or organizations?
4. Is there a path to monetization, and what is your go-to-market strategy?
5. What are the key challenges in scaling this platform beyond one developer’s prototype?
6. How do you ensure that AI-generated content meets safety standards before use?
7. Are there any existing partnerships with educational institutions or safety organizations?