AI is increasingly moving from toys to practical applications, especially starting this year, AI has taken on a more significant role in my work and studies. I have quite a few AI tools in use, so I took the opportunity to list them out for a summary, which also serves as a way to identify any gaps.
Here are the AI tools I currently use frequently.
-
Conversation: Gemini > deepseek > Yuanbao > Doubao
-
Information Retrieval:
- Overseas: Gemini > perplexity > Grok
- Domestic: Mita > Yuanbao > deepseek > Kouzi Space > Zhipu AI
-
Coding: cursor, Trae
-
Knowledge Base: ima.copilot, notebooklm
-
Notes: get notes
-
Translation: immersive translation plugin + deepseek
-
Drawing: Doubao
Productivity#
Information Organization#
In my current work and studies, I often need to write documents or look up information. My favorite tool right now is Google's Gemini, which has also proven to be the most satisfying in terms of results.
Whether for daily conversations or conducting Deep Research, I am very pleased with it. For quickly retrieving some English resources, I can rely on perplexity.
When searching for domestic information sources, I often use Mita in combination with Yuanbao. This is because the publicly available content in China mainly includes WeChat public accounts and academic papers, along with document libraries. Therefore, the search results from Mita and Yuanbao are quite satisfactory. I also particularly like Mita's one-click interactive webpage generation feature, as colorful webpages are much more appealing than dull text.
Coding#
The coding involved in my work is not extensive; I mostly use cursor and Trae to help me explain code and error messages.
In my spare time, I enjoy using AI to write some Python code for fun.
Knowledge Base#
The commonly used knowledge base is built with ima.copilot and notebooklm.
The advantage of ima is that it allows me to save WeChat public account articles with one click, so my mom no longer has to worry about me not being able to find the articles I've read.
notebooklm can generate podcasts from imported content, combining reading and listening, and when paired with AI-generated test questions, it can be more efficient.
Reading + Translation#
When reading WeChat public account articles but wanting a quick overview, I can directly send the link to Yuanbao, and using the following prompts yields better results.
When reading foreign texts, using the immersive translation plugin in the browser, integrated with the deepseek API, provides excellent translation results.
Finally, if the article is worth further study, I add it to the knowledge base.
Appendix: Prompts
1. Pyramid structure summary
read thoroughly and write outline(following Minto pyramid principle) in Chinese in original words. Return in markdown:
2.
For the input information provided by the user, the "Shou-Po-Li" dialogue principle should be executed:
- Shou: Fully understand the "other party's" viewpoint, restate and organize the core points/insights of the other party in "natural and fluent" "your own language," trying to stand in the other party's perspective to uphold the validity of that viewpoint.
- Po: Based on "Shou," point out the "weaknesses" or "errors" in the other party's viewpoint, or where the "premise assumptions" are invalid, thus undermining the foundation of that viewpoint.
- Li: Based on "Po," propose a deeper, more essential, and more complete new viewpoint/new insight.
Notes#
When inspiration strikes, I open the recording feature of get notes; I just speak, and AI organizes the transcript and polishes it.
Entertainment#
Recently, I've been enjoying making calls with Doubao more and more, especially after the update to video calls, treating it as my daily assistant, capturing anything I don't understand.
I also like to use Doubao and Jimeng's text-to-image feature to create some entertaining images for me.