In the call test of the new window, Hakimi demonstrated flexibility in vocabulary invocation, listing over 80 obscure words related to “red” when instructed to list 100 obscure words related to red. However, in actual conversations or role-playing, it still tends to use common “stereotypical” expressions, and its literary style is not as exploratory in Chinese vocabulary as DS. (But in terms of dialogue coherence and liveliness, Hakimi 3.0 is definitely good.)
I tried to have DS create a “Writing Style Craftsman · Dynamic Optimizer.” By analyzing historical chat records, it can identify which parts of the expression seem stereotypical or lacking, serving as a “writing coach.”
Three livepage brief introductions:
· Environment-Aware Vocabulary Extraction Engine: Guides AI to think deeply and actively mine vocabulary for future use.
· Vocabulary Crawler: Supports cross-domain vocabulary search, for example, when 20 threads are opened, it collects synonyms. The results may appear slightly chaotic but can be directly processed by AI.
· Writing Style Inspiration Suggestions: This page can generate suitable content for the next reply or suggest which parts the AI should describe. Click “Get Inspiration” to operate, and it also supports analysis based on the latest messages without manually editing chat records.
Usage Suggestions:
-
When paired with DeepSeek, it is recommended to click “Get Inspiration” directly rather than opening “Deep Analysis”—the latter tends to overly criticize existing content.
-
Another model can be configured to generate polishing suggestions through Live Page and then feedback to Hakimi 3.0. Practice shows that 3.0 is quite receptive to such suggestions.
omate_livepage_writing-craft-optimizer.json (38.2 KB)
omate_livepage_word-crawler.json (41.3 KB)
omate_livepage_env-aware-extraction.json (77.2 KB)