Extra! Extra! Today marks the launch of this year's first vibe coding project: KOL Lens.
- Why am I doing this?
Twitter is the most important source of information in the crypto space. Therefore, my plan this year is to find a list of high-quality KOLs and conduct in-depth analysis using AI.
The targets mainly include:
1) KOLs skilled in analyzing BTC trends
2) KOLs adept at uncovering early Alpha
3) KOLs proficient in on-chain data analysis
However, I encountered a major problem: how to assess a KOL's "expertise"?
Currently, KOLs are numerous, and I need to identify truly high-quality accounts among them. I researched tools like Kaito, Xhunt, and Cookie, but found that none fully met my needs. Manual review is not only time-consuming but also difficult to ensure accuracy.
Thus, with my background in data analysis and the support of vibe coding, I decided to create a comprehensive data analysis tool for KOLs myself.
This time, I analyzed over 30,000 KOLs based on nearly 9 million tweets. Fortunately, I recently put significant effort into fine-tuning a model specifically for analyzing the crypto space; otherwise, using the Gemini-3 API for analysis would have cost nearly $100,000 just for a round of entity recognition.
Below is an introduction to various data dimensions:
- Basic Situation of Twitter
When selecting KOLs, the first data point to look at is the average number of tweets per day.
If the number is too high, such as more than 8 tweets per day, there is a high probability that AI writing is continuously posting. Even if a person is writing, such high-frequency content often has a low proportion of effective information. Conversely, if the number is too low, it indicates inactivity, and there is no need to follow.
Next is the distribution of tweet length. If long and extra-long tweets account for a large proportion, it indicates that the blogger prefers to produce in-depth content. In the screenshot, the proportion of short tweets is quite high, with fewer words.
Combining this with the distribution of views, the account in the screenshot has tweet views generally between 1K and 5K.
Thus, the basic profile of this KOL has been sketched out: this is an account that tweets a lot daily (more than 5 tweets), but the content is very short, and the view count is low. If it were me, I would pass immediately.
Additionally, by analyzing the tweeting times across different periods, we can see its activity patterns. For example, the screenshot shows that the highest number of tweets occurs at 7 AM UTC, likely indicating scheduled posts. The periods with very low tweet counts should correspond to sleeping hours.
It can be inferred that this account is not based in the U.S., and its sleeping time is likely around 7 AM Beijing time.
- Analysis in the Blockchain Aspect
First, I set two indicators to analyze the "crypto content" in KOL tweets:
Average tweet length: As seen in the above image, the word count for blockchain-related content (444) is significantly higher than for non-blockchain content (142), indicating that this KOL puts more effort into writing about crypto topics.
Proportion of blockchain tweets: The screenshot shows 57%. This means that 57% of the tweets are about blockchain. This figure is too low and is a dealbreaker for me. If it were below 10%, it would mean that 90% of the tweets are not about blockchain, suggesting that this person's main focus is not on crypto.
Secondly, more important information can be gleaned from the chart:
Number of discussed coins: The screenshot shows 393. Clearly, this is too many; this KOL is calling out a lot of projects. Even if some coins have high returns, the probability of making money by following them is low.
Distribution of token market caps: This can easily reveal the KOL's preferences. Do they prefer large-cap or small-cap tokens? This is an extremely important indicator when looking for different types of KOLs. If searching for a BTC trend expert, a high proportion of large-cap tokens is essential; if looking for early Alpha KOLs, a very high proportion of small-cap tokens is necessary.
Distribution of preferred sectors: It is immediately apparent which fields this KOL discusses projects in. For example, the x402 sector in the image has 11 projects, indicating that they are closely following trends.
From the screenshot, it can be seen that this KOL pursues early Alpha and keeps up with trends. However, the problem is that the number of analyzed projects is too high, making them not the ideal candidate, so they can only be placed on the pending list.
- PNL (Profit and Loss Analysis)
The KOL's track record is the most direct indicator of their level.
Looking at the KOL in the screenshot, in the past 3 months, they have had 2 coins with 20x returns and one with over 10x returns, which is indeed impressive. I have already added them to the selected KOL list, hahaha.
In addition to looking at direct data, there are other angles that can reveal insights:
Listed tokens: If they are all small-cap, it indicates that this KOL is quite skilled at identifying early Alpha projects.
7-day max ROI vs 30-day max ROI: If the two values are the same, it indicates that the KOL's recommendations reached maximum returns within 7 days, suitable for short-term operations. If the 7-day max ROI is small but suddenly increases at 30 days, it suggests that the projects recommended by this KOL are suitable for long-term holding.
Of course, PNL can uncover more information, and I plan to continue improving it in the future.
- Website Information
Currently, https://riyuexiaochu.vip/ is accessible. In the website's search bar, various KOL data can be viewed.
Considering that the early website functions may still be incomplete, a limited testing phase is in place, requiring an invitation code to register and use.
Friends who need an invitation code can leave a message here.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
