AI-Driven Development based on Natural Language framework_VercelMeetupLT#2

AI-Driven Development based on Natural Language framework_VercelMeetupLT#2

1. AI-Driven Development based on Natural Language framework Vercel Meetup #2 AI and v0 special with CEO

2. Self introduction 　 Rinte Developer of GEAR.indigo GEAR.indigo is a product that can generate development document and source code EN:@rinte2409 JP :@rinte0321

3. What is Natural Language framework Natural Language framework is information contained in development documents such as requirements definition, design documents. Operation flow System overview Functions list Screen transition Screens list System architecture Table definition Sequence diagram

4. Cases you should develop with using Natural Language framework The Natural Language framework provides a base for project, making team development and large-scale development easier Current system generation AI focuses on strengthening individuals. Natural Language framework can be a method to support team development from start to finish. Team development When building a large-scale system, the connections between the frontend and backend, and between screens, become complicated. By using the Natural Language framework as a base, it can scale with correct connections. Large-scale development

5. Positioning of Natural Language framework Natural Language frameworks have strengths in early system design and coding Coding Deploy System design Natural Language framework High Quality Small Amount Low Quality Large Amount

6. Steps of the development using Natural Language framework The steps are to first create a base with a natural language framework, then supplement it with components, and then modify it with an editor. Product Natural Language framework products v0 Claude ChatGPT Gemini Cursor Codeium Github copilot Generate base of project Generate components Minor corrections Step 1 Step 2 Step3 Documents Base code Documents Base code Com ponent Documents Base code Minor corrections Com ponent Com ponent Com ponent Com ponent Com ponent

7. Products using Natural Language framework 　

8. About Babel & Zoltraak Babel & Zoltraak consists of a Natural Language framework and an editor, with the Natural Language framework running on the backend

9. About Babel & Zoltraak – yaml file By defining the order in which files are generated in a yaml file, the system can be generated while maintaining the correct connections between files

10. About GEAR.indigo – Generate Documents A is a tool that can generate development documents and source code. It adds elements that assist AI generation based on documents that are easy for humans to interpret

11. About GEAR.indigo –Strategy of generating documents and source code When generating the documents, the connections between frontend and backend are defined in each documents and these are used for generating source code. Generate frontend Generate backend Including the relevant frontend code into the prompt Generate docs item item item item Generate a list of screens and define the configuration, operations, and file names for each screen. Generate a list of backends and define related screens, operations, and file names.

12. Improvement Because the output depends on the performance of the LLM, we want to use the more powerful LLM v0 via the API. The output code of v0 has fewer errors, making it easier to achieve immediate deployment. Current Problem Each output depends on the performance of the LLM. Currently available LLMs have excellent performance, but there are issues in the design area, such as drawing architectures and flow diagrams. Solution I want to use the v0 API because v0 outputs beautiful UI and clean frontend implementations. I want to use it to generate the following: Diagrams Wireframes Frontends Ideal Flow v0 can output code with few errors and excellent design, so the moment it is output, it can be linked to the Supabase and deployed to Vercel. I think this matches with Vercel's vision, so I hope Vercel will consider it.

13. Summary There are many products for system generation Natural Language frameworks are effective for team and large-scale development It creates a base so it can be expanded By linking with products such as v0, Claude, and Cursor, it is possible to provide a higher quality product development experience Among them, v0's front-end implementation and design capabilities are outstanding I would love to see v0's API opened

14. EoF