It is the morning of January 10, 2026.
The refactoring of CZON's JSX rendering has been preliminarily completed. Following the approach outlined in this document, CZON can now render different types of pages using JSX components. The original AI-crafted template engine based on Placeholder-Replacement has been removed and replaced with JSX from the React ecosystem for page rendering. However, the custom theme feature has been temporarily put on hold to focus on perfecting the core functionality.
As the text content gradually becomes richer, it seems about time to start building the AI classification feature, which is one of CZON's core selling points. I don't want to manually categorize the content, especially since the volume is growing, making manual classification cumbersome. Therefore, I plan to let AI assist with this task.
In simple terms, I intend to first have the AI read all the content in .czon/meta.json, then generate a classification system based on that content, and finally have the AI assign classification tags to each article according to this system. This will enable automatic classification.
Clearly, this stage must occur after the metadata for individual articles has been extracted but before the enhancement of Markdown metadata. This is because it may generate some classification tags that need translation, so it must be completed before generating the .czon/src native language Markdown files.
When performing AI classification, it's best not to generate too many categories, as that would make things appear cluttered. Each category should contain at least three articles; otherwise, the categories would become too fragmented. This principle is similar to the idea that code should be repeated at least three times before being recommended for extraction into a component.
Additionally, the number of categories should be limited, preferably to no more than ten. If there are too many, users may feel overwhelmed when browsing the classification directory, which would hinder content discovery rather than help it.
Of course, each article can only belong to one category, avoiding multiple classifications. This ensures clarity and simplicity in the classification system.
Once classification tags are in place, features like categorized browsing and recommendations can be implemented later. Users will be able to browse related content based on classification tags, and the recommendation system can prioritize articles from the same category, thereby enhancing the user's reading experience.
Time to get started, let's go!
