In 2023 and 2024, our main focus is on improving UI restoration accuracy and better component recognition, and we have accumulated a number of users.
With the improvement of AI model capabilities in 2024, more companies and teams are beginning to explore application development through AI, similar to v0.dev, bolt.new, etc. However, when it comes to UI generation, Text2Code or Image2Code methods make it difficult to get precise UI results, and it's hard to get satisfactory results compared to simple demos and vague requirements.
It is expected that in the next 1-2 years, the proportion of development based on design drafts will remain high, and the demand for precise UI results will not be eliminated.
At this time, D2C results can be used to help large models accurately understand design drafts as an intermediate link in AI application development. On the other hand, AI's code and image understanding capabilities can also help D2C improve intermediate links and downstream processing links, such as improving the maintainability of delivered code, responsive layout capabilities, and supplementing business interaction logic.
Therefore, in addition to improving basic capabilities in the future, we will make more breakthroughs in the combination of D2C and AI capabilities.
- At the core capability layer, the current engineering-based approach to UI restoration improvement is no longer obvious. In the future, we will introduce AI to understand images and Figma raw data to help generate more reasonable Dom structures and responsive layouts.
- At the component access layer, the current cost of user access to component recognition is too high, hindering the expansion of D2C user scale. In the future, we will try to achieve component recognition by reducing component marking costs or even abandoning component marking.
- At the consumption pipeline layer, on one hand, we will optimize code maintainability through the D2C platform and AI rewriting capabilities. On the other hand, explore how to better consume and understand D2C code by large models, and explore business logic generation, data access and deployment capabilities in certain areas with specific specifications.