X's open source this time has shown us the underwear of the recommendation algorithm. The core of the recommendation algorithm is directly installed with the Grok large model. The era of relying on engineers to manually adjust parameters in the past is over. I used to study what tags could make me popular or when to post with high weight, but now these tricks are all ineffective. The two main cores of algorithms This system has two very interesting code names. One is called Thunder, which means Thunder. It specifically manages the acquaintances you follow, ensuring that you can quickly catch any updates posted by people in your circle. The other one is called Phoenix, which means Phoenix. It is used to capture content across all platforms that you may not be following but may be interested in, matching your preferences with global tweets. The current recommendation logic is very simple and crude, focusing on the real behavior of each user. Who have you liked and commented on, it will predict fifteen possible reactions you may have after seeing a piece of content. The true value of interactive weights Many people think that liking is important, but in the eyes of algorithms, liking is almost the cheapest interaction. The weight of likes may only be around 0.5. The weight of dialogue interaction is surprisingly high at 75. If you post a piece of content and someone comments below, and then reply to them with a single sentence, this kind of interactive response can generate more than 100 times the weight of likes. What algorithms most want to see now is real conversations in the community. A one-way dot with a small red heart no longer holds much meaning. The platform always holds the right to distribute traffic The platform always holds the right to distribute traffic. Even if fans follow you, they may not necessarily see your messages. After reaching a certain limit, the effect of increasing exposure by fans will significantly decrease. Prevent you from seeing too much content from the same author. That's also why sometimes we find that even though we follow a certain KOL, we always feel like we can't get their full information. A negative signal of immense destructive power The crackdown on fake interactive platforms is now severe. The original KOLs may buy some zombie fans or go to various mutual aid groups for mutual evaluation in order to gain followers. This road is basically closed now. The algorithm will now calculate a total score. There are not only additional items here, but also extremely terrifying deduction items. If someone accidentally blocks or reports something you post because it's too noisy or too watery, the negative signal's destructive power is astonishing. A block may require you to use over a hundred likes to offset it. If your account is hacked by many people, your credit score will hit rock bottom, and almost no one will read anything you post. The Power Reduction Trap of Diligent Bloggers The more posts there are, the greater the traffic. There is a mechanism hidden in the code called the Author Diversity Scorer. Its logic is very simple. The system does not want the same person's content to occupy all of the user's time. If you send several consecutive messages in a short period of time, the algorithm will automatically give discounts to the following content. If your first post is full marks, by the time you reach the fifth post, you may only have 20% of the score left. Frequent screen scrolling does not increase reading volume. A guide to avoiding pitfalls in external links and videos Friends in the Web3 community often like to post long articles or include links. I used to think that external links would reduce authority, but now the code confirms this. Algorithms do not like to take people out of the station. If you really want to attract traffic to your project or blog, the best way is to hide the link in the first comment section. Leave the main text for those words that can evoke resonance and discussion. There is also a counterintuitive detail about the video content. Although Musk has been advocating for video first, the weight of the new algorithm on video completion rate is actually absurdly low. It doesn't regard the completion rate as the root of success like Tiktok. You don't have to force yourself into a short video director, bury a dot in a few seconds. Focusing on producing profound text and images remains the main theme of Twitter. The only way to break the game for a small account If you are a newcomer to the industry, don't always rely on the platform to automatically push you streams. The discovery flow of the new algorithm is not friendly to small accounts. You should first interact with influencers and leave high-quality comments under those popular posts. If your comment is liked or replied to by a big V, the exposure effect it brings is much stronger than talking to yourself. summary Nowadays, X is becoming more and more like a smart creature. It can feel that you are sincerely sharing your opinions or mechanically engaging in marketing. Instead of constantly researching how to use conventional algorithms, it's better to look back and think about what value your content has brought to the community. Sincerity is always the biggest traffic password on this platform. Go have real conversations with your fans and create those warm collisions. Traffic will naturally follow you.