Original video in Chinese.
Key Takeaway
- AI analyst capabilities: Claude Skills created Super Analyst to handle complex strategic assessments (such as ChatGPT Atlas’s chances of winning), with an initial evaluation of Level 3, intelligence gathering (12 rounds of thinking/8 rounds of search), and multi-framework analysis (Porter’s Five Forces/SWOT/scenario analysis, from outside in / from present to future logic).
- Pro version upgrade: integrated the Sequential Thinking MCP tool, and established a 7-step SOP, including problem judgment (type/complexity/need for external information), intelligent intelligence planning (information needs/priority/keywords/strategy), framework selection/application, improving deep thinking/research/structure.
- Comparison and sharing: big differences in the results of the native/basic/Pro versions (from simple summaries to deep analysis); the community shares Skill files/conversation links/comparisons; it’s easy to make (just paste the requirements to Claude), and you’re encouraged to modify it into a personalized capability package.
My AI analyst is definitely more powerful than most human analysts. It can do deep thinking, extensive research, and structured analysis.
For example: OpenAI recently launched ChatGPT Atlas. I asked the AI analyst to help me analyze this product’s chances of success.
Once it receives the request, it first makes an initial assessment and finds that the question is fairly complex, a Level 3 strategic evaluation problem.
Next, it develops an intelligence-gathering strategy. You can see that it went through 12 rounds of thinking, clearly identifying the need to gather product information, competitor information, technical information, as well as media and user feedback, plus the number of searches, the priority order of the information, and so on. In the end, it formed a keyword combination covering every direction.
After 8 rounds of searching, the AI had enough information. It then moved on to the next step: choosing the analysis framework.
Different problems require different analytical frameworks. For the ChatGPT Atlas question, the AI considered two things:
First, the essence of the problem is a strategic evaluation problem.
Second, it had already gathered multi-dimensional information, enough to support multi-framework analysis.
So, the AI decided to use three frameworks: Porter’s Five Forces, SWOT, and scenario analysis. And in terms of the order of use, it decided to follow an “outside in, present to future” logic, starting with Porter’s Five Forces, then SWOT, and finally scenario analysis.
With such thorough preparation, I was very satisfied with the final output. Think about how much time a human analyst would need to produce an analysis like this.
The AI analyst I used is actually a capability package made with Claude Skills. I shared it in the previous video. But yesterday, I upgraded it and greatly improved its deep thinking ability. In this video, I’ll introduce it in detail.
Also, I’ve shared the Skill file in the community, so everyone can grab it themselves.
Hello everyone, welcome back to my channel. Humbly speaking, I’m one of the few bloggers in China who can clearly explain the Why and How of AI. What I offer is far more valuable than tutorials. Remember to follow. If you want to connect with me, come to our newtype community. The community has been running for 600 days, and more than 1,700 friends have paid to join.
If you’re a domestic user, you can join through Knowledge Planet. If you’re overseas, you can join through Substack. My first course, daily newsletter, and exclusive videos can all be found inside the community.
Back to today’s topic: Super Analyst Pro.
As for the AI analyst, I made a capability package through Claude Skills a few days ago, which is the Super Analyst basic version. Its core function is to provide the AI with 12 analysis frameworks and help the AI choose among them.
But this basic version did not meet my requirements. I believe an AI analyst should have three capabilities:
First, it should be able to make an initial judgment about the problem. Initial judgment is important because it affects how many resources will need to be invested later. We can’t just keep going and see, right?
Second, in terms of intelligence gathering, it should be able to develop a complete search strategy based on the problem. Right now, most AIs are very casual about picking a few keywords and just starting; that’s not professional enough.
Third, in terms of framework use, after selecting multiple frameworks, how should those frameworks be combined and used? There has to be a logical order, right?
All three of these capabilities point to a deeper underlying ability: deep, multi-step thinking.
Then I thought of an MCP I especially like: Sequential Thinking.
I won’t go into too much detail about this MCP. After MCP became popular at the beginning of this year, I recommended it in many video episodes.
This time, I decided to call it into the Skill as well.
Before formally starting, I first asked the AI to fully understand the status of the Super Analyst basic version. That way, we’d actually have something to talk about later.
After it finished everything, I told it the full requirement description and asked it not to rush into implementation, but to first provide a plan. We would discuss it thoroughly before starting work.
Sonnet 4.5 is actually pretty powerful. It understood exactly what I meant. It believed that Super Analyst Pro could perform deep thinking, extensive research, and structured analysis. One of the main upgrades was to use Sequential Thinking for intelligence planning and framework selection.
Based on that understanding, the AI developed a 7-step SOP.
When it receives my analysis request, it first judges the problem’s type, complexity, timeliness, and so on. One key judgment is whether external information is needed.
If it is needed, it will activate the “intelligent intelligence planning” capability and first think about:
What specific information does this problem require? For example, quantitative data, qualitative information, or background knowledge?
What are the priority and dependency relationships among this information? Which is essential? Which is supplementary? What should be searched first, and what should be searched later?
What is the best keyword design for obtaining this information? Chinese keywords or English keywords?
And what is the specific search strategy plan? How many rounds of searching are needed? What should each round focus on?
You see, just this intelligence-planning capability already surpasses most humans.
Likewise, when it gets to the framework-selection stage, with the help of the external tool Sequential Thinking, the AI will think about: what is the essence of the problem? What characteristics does it have? Which frameworks are most suitable? And what is the strategy for applying the frameworks?
Once all of the above preparation work is complete, the AI then enters the structured execution stage. It will fully combine the intelligence gathered through search, complete the analysis according to the framework steps, and then produce the output.
I did a comparison: for the same question, one version did not use MCP or Skills and only had native capabilities; one used the Super Analyst basic version; and one used the latest Super Analyst Pro. The differences in the results generated by these three were quite large.
I’ve posted all three conversation links in the community. Also, for this upgrade, I shared my conversation link with Claude in the community too. Everyone can fully see how a Skill is made. It really isn’t hard.
Actually, after you get the Skill file I shared, you can paste it to Claude and make further modifications based on this Skill, turning it into a personalized capability package of your own. Claude can absolutely help you do that. So please be sure to give it a try.
OK, that’s all for this episode. If you want to learn about AI, want to become a super individual, and want to find like-minded people, come join our newtype community. See you next episode!