Original video in Chinese.
Key Takeaway
- Gemini and Grok are two underrated top-tier AI apps that show strong productivity in different ways.
- Gemini offers a versatile experience, including quick answers (Flash), article writing (Canvas), and deep reports (Deep Research), and integrates seamlessly with Google Docs and NotebookLM to support deep learning.
- Grok feels more “human” in its model responses, has a simple product design, and its unique Twitter data source gives it an advantage in gathering real user feedback.
- Using Gemini together with Grok — Gemini for mainstream media information and Grok for social media information — can produce more comprehensive and realistic insights.
Full Content
There are still two top-tier AI apps that people in China are overlooking. One is Gemini, and the other is Grok.
I’ve already recommended Gemini enough. Look, over the past six months, I’ve really been persuading people and repeating this in the community over and over. I started with the 1.5 model and moved on to 2.0 and 2.5; the more I use it, the more I like it. Forget the complicated stuff — just speaking about everyday use, like I said in this post:
If you just want a quick answer, ask Gemini 2.0 Flash directly. It’s fast, strong at reasoning, and runs stably.
If you want to go a little deeper, then turn on Canvas. Gemini 2.0 Flash will answer you according to the standards of article writing.
If you want a detailed report, turn on Deep Research. Gemini 2.0 Flash will output it according to report standards.
Let me give you a demo directly.
For example, I ask Gemini: what is a Reasoning model? It quickly gives a brief answer.
I turn on Canvas, which is the drawing board. I ask it the same question again. At this point, the response is a little slower, but the level of detail is much higher. In Canvas, you can format and edit the text, and you can also select a section and ask a question about it. In other words, with this feature, you can work with Gemini to complete the creation of an article.
If I turn on Deep Research? It’s still the same question — what is a Reasoning model. Gemini will help me break down the topic, search hundreds of web pages, and ultimately produce a deep report.
You see, it’s that simple: three functions — direct conversation, Canvas, and Deep Research — cover the main scenarios in our daily use of AI.
Going one step further, if you think these answers are valuable, you can also save them to Google Docs.
Of the three functions I just introduced, the latter two can both be exported to Google Docs. Then, just like I introduced in my previous video, open NotebookLM, import all the relevant content from Google Docs, and then you can carry out deep learning.
Gemini has other features too, like image generation, and so on, but I won’t say much more. Just what I’ve introduced so far is already worth the price.
And Google keeps updating it continuously. Just yesterday, they launched the first hybrid reasoning model — Gemini 2.5 Flash. You can choose whether to turn on thinking mode, and you can even set how many resources to invest in thinking. The arrival of this model further strengthens my confidence that I should keep embracing Gemini.
So since Gemini is so damn good, why do we still need Grok? Three reasons:
First, in terms of how the model answers feel, Grok 3 is much better than Gemini 2.0.
Gemini 2.0 still has a kind of stiffness and mechanical feel to it — maybe that’s just Google’s own flavor. But Grok 3’s responses feel more “human,” so I’m gradually leaning toward talking to it more.
Second, from a product perspective, Grok is the most complete.
Like I said in the community before, Grok doesn’t make users choose from a long list of models. There are just two buttons: deep thinking and deep search. There’s no distinction between models, only between modes, and later on you won’t even need to choose modes — the AI will judge on its own. That is the right path. Something like OpenAI’s approach is a bit insane.
Also, the update from the other day added Canvas and Workspace to Grok. Canvas is pretty much the same as everyone else’s — a standard drawing board. Workspace, meanwhile, is similar to a knowledge base feature, and it’s especially useful.
Now Grok has basically gathered all the mainstream features available right now. On the product side, it’s already quite mature.
Third, from the perspective of data sources, Grok has Twitter data that no one else has.
Twitter is now the most influential social media platform, and its data is very important for AI. If you can combine the social media data represented by Twitter with the mainstream media data Google has access to, it will definitely help improve the quality of AI responses. How exactly do you do that? Let me show you again.
Actually, I’ve shared this method in the community before too. It’s simple: besides supporting local document uploads, Grok’s Workspace feature also supports importing from Google Drive — that’s the key function.
I’ll first use Gemini to do a round of Deep Research, then import the report into Google Docs. After that, I add these Google Docs into Grok, and in the prompt I emphasize: the attached information is only for reference; you need to do your own searching and judgment. Also, what feedback are users giving on Twitter? Focus on that.
Pay attention to Grok’s analysis process. In terms of collecting information from mainstream media, it really doesn’t find as much as Gemini does. Actually, it’s not just it — OpenAI’s search scope is limited too. But Grok can collect data from Twitter, including posts and comments, and that is its unique value.
I spent more than ten years in the PR and marketing industry, and I know very well that as long as your budget is enough, most mainstream media content can be bought, or at least influenced. By comparison, users’ actual usage feedback is much more real. So when the topic I want to understand involves user reputation, I’ll definitely use Grok to collect it, and then combine it with the mainstream media information from Gemini.
The usage experience of these two AI tools I just introduced may seem really ordinary, not as flashy as the tricks many AI influencers talk about, but these are genuinely useful things that can help you. If you think they’re valuable and want to learn more, remember to join our community.
OK, that’s it for this episode. See you next time!