The Evil of Algorithms: From Twitter to Musk's X, Why You Feel the Quality Has Declined

Many people may feel that the current X, compared to the Twitter era, has deteriorated in content quality. This feeling is even more pronounced among those in the Chinese-speaking community. While the CCP's intensified efforts to influence overseas public opinion contribute to this, the more fundamental reason is Musk's complete destruction of Twitter's underlying logic.

In the past, Twitter's content discovery primarily relied on posts and retweets from people you followed. Therefore, to gain followers, you had to pay close attention to the quality of the content you posted and retweeted. High-quality content was filtered through the number of followers and retweets. It can be said that this was a large-scale human filtering mechanism. Now, X has co-opted traffic, and it uses its own algorithms to determine what content is high-quality and forcefully recommends it to more people. How can such algorithmic judgment surpass human judgment?

As a result, everyone sees that as long as you deeply understand X's recommendation algorithm and cater to it, you can gain higher readership. This, of course, lowers the overall content quality of X. Or rather, it lowers the overall quality of the X content you can see.

Below is AI's answer to this observation 1: You've actually hit on a core issue: a platform's "distribution logic" change directly reshapes the content ecosystem.

In the Twitter era, the main path for content dissemination was indeed "following relationships + retweet diffusion." This meant:

• You want to be seen → You need to be actively followed
• You want to be followed long-term → You need to consistently produce valuable content
• Retweets served as a "seal of approval"

Essentially, it was a social graph filtering mechanism.

In the X era, the platform has strengthened the "algorithmic recommendation feed (For You)." Content is no longer primarily spread through people you follow but is forcefully pushed to you by the system predicting what "you might be interested in."

This has brought about several structural changes:

1. From "Relationship-Driven" to "Attention-Driven"

Old logic:

Relationship → Trust → Dissemination

New logic:

Engagement Rate → Algorithmic Amplification → More Engagement → Further Amplification

This is a typical "feedback loop reinforcement mechanism." The algorithm doesn't judge "truth" or "depth"; it judges:

• Dwell time
• Like rate
• Comment rate
• Retweet rate
• Provocation level

Consequently, content that is highly emotional, conflict-driven, or extreme is naturally amplified more easily.

1. Algorithms Cannot Truly Understand "Quality"

You said, "How can algorithms surpass human judgment?" There's a key difference here:

• Humans judge quality by considering long-term credibility, logical coherence, and historical consistency.
• Algorithms judge quality based on short-term behavioral data models.

Algorithms make "statistical judgments," not "value judgments."

If an account understands how to:

• Extreme complex issues
• Provoke opposition with strong emotions
• Create controversial headlines
• Time posts precisely
• Engage frequently to boost metrics

It can capitalize on traffic bonuses.

This doesn't equate to "better content" but rather "more aligned with the algorithmic amplification mechanism."

1. Why is the Chinese-speaking community's feeling more pronounced?

You mentioned a point: the Chinese-speaking community's perception is stronger.

This could be due to several reasons:

1. The overall volume of Chinese users is smaller than English users.
2. Content is more easily manipulated organizationally (by any party).
3. The algorithm's semantic understanding in the Chinese context is weaker.
4. Extreme expressions can more rapidly form "information cocoons" in smaller language communities.

When the social graph is weakened, the small-circle structures maintained by "long-term follower relationships" are broken down.

1. Does this count as "destruction of underlying logic"?

In a sense, you are right—

The platform has shifted from a "social network" to a "content platform."

The Twitter era was closer to:

A distributed social relationship network

X is closer to:

An algorithm-centric content distribution machine

This isn't simply "getting worse"; the objective function has changed.

Previously, it optimized for "relationship activity." Now, it optimizes for "overall interaction volume and platform activity."

1. A practical problem

Even if the algorithm amplifies low-quality content, why does the platform do it?

Because:

• Emotion is more contagious than rationality.
• Conflict leads to longer dwell times than depth.
• Extreme expressions generate more engagement than moderation.

The algorithm isn't "judging incorrectly"; it's that its optimization goal is not the same metric as the "quality" in your mind.

---

If your statement were compressed into a more theoretical expression, it could be said:

X's problem is not that the content has worsened, but that content distribution has shifted from a "human screening mechanism based on social trust" to an "algorithmic amplification mechanism based on interaction data," leading to a structural advantage for emotional content. Below is AI's answer to this observation 2: You've articulated it very clearly. The phenomenon you've observed is essentially the growing pains of social media transitioning from a "social graph-driven" model to an "interest algorithm-driven" model.

In the Chinese-speaking community, this "perceived deterioration" is particularly intense. From the perspective of the platform's underlying logic, let's break down why Musk's changes have made many deep users feel that "Twitter is dead":

1. From "Human Endorsement" to "Weight Game"

As you said, the core of old Twitter was following.

Old logic: A tweet became popular because many real people you trusted and who had good taste manually clicked "retweet." This was a distributed filtering process involving human intelligence and aesthetics. New logic: X's algorithm prioritizes dwell time, engagement rate (even if it's abuse), and membership weight (Blue Check). If you purchase Premium, your comments are forcibly pinned to the top. This leads to "money buying voice" rather than "quality earning voice."

1. The Algorithm's Natural Preference for "Controversy"

Algorithms are inherently declining. They don't understand "depth"; they only understand "stimulation."

• In the Chinese-speaking community, this manifests as a large number of marketing accounts, content scrapers driving traffic, and highly inflammatory emotional remarks (even AI-generated nonsense).
• Because these types of content most effectively provoke rebuttal, debate, or clicks, the algorithm mistakenly believes they are "high-quality content" and pushes them to your "For You" page.
• Consequence: Serious discussions are drowned out, and attention-grabbing "traffic codes" dominate the information flow.

1. The Special Predicament of the Chinese-speaking Community: Invasion of Spam

Musk's layoffs of numerous teams responsible for content moderation and community governance have created opportunities for specific forces:

• Gray industries and political bots: The current Chinese X is flooded with numerous pornographic lead-generation bots and organized political bots.
• Ineffective countermeasures: Previously, real user reports and credit weights could filter out this noise, but under the current algorithmic logic, as long as these accounts interact with each other (liking, commenting), the algorithm considers them an "active community" and recommends them to real users.

Summary

Your statement "lowered the overall quality of the content you can see" is very accurate. X is no longer an "information filter" but has become an "attention black hole." It no longer serves to help you obtain valuable information but rather to keep you on the platform longer.

For deep users in the Chinese-speaking community, this change is catastrophic because we not only have to face the mediocrity of the algorithm but also more complex propaganda offensives. How to improve? With the underlying logic of X (Twitter) fundamentally changed, we can no longer passively wait for the algorithm to "feed us" as before. Instead, we must regain control of the information flow through manual adjustments.

Here are some advanced techniques tailored to the current situation in the Chinese-speaking community (prevalence of marketing accounts, bots, and spam ads) that can effectively help you turn X back into the efficient information acquisition tool it once was:

1. Forcefully Switch to the "Following" Tab

This is the most direct way to combat the algorithm.

• Operation: Always click "Following" at the top of the screen, not the default "For You."
• Tip: X often switches back to "For You" when you reopen the app. Make it a habit to immediately switch back if you see messy recommendations. Only here will you see the information flow you personally selected.

1. Use "Lists" to Build Clean Niches

If you find "Following" still too messy, Lists are the ultimate solution.

• Deep Filtering: Create several private lists (e.g., Tech, Politics, Interesting People).
• Disturbance-Free: Put the high-quality bloggers you truly want to follow into these lists. Most importantly, content within lists is strictly sorted by timeline, with no algorithmic recommendations or ad insertions.
• Pinning: You can pin lists, allowing you to switch between different "sanctuaries" by swiping left and right on the main page.

1. "Block Keywords": Precisely Target Marketing Accounts

Currently,

There are many commonalities in Chinese spam messages. Using the "Muted Words" feature can significantly reduce interference:

• Muted Words Library Suggestions:

• Traffic diversion: "See comments section", "Uncensored", "Direct link", "Pinned".

• Gray industry: Block specific adult or gambling keywords.

• AI nonsense: Block words that frequently appear in AI-generated content, such as "very meaningful", "In summary", "have to say".

• Path: Settings and Support -> Settings and Privacy -> Privacy and Security -> Mute and Block -> Muted Words.

1. Enable "Quality Filter"

Although algorithms cause many problems, this native filter is effective in filtering out low-level bots.

• Function: It automatically hides duplicate tweets and suspected automated spam replies.
• Path: Notifications -> Filters -> Enable "Quality Filter".
• Advanced: On the same page, go to "Muted Notifications" and check "Accounts with unconfirmed email addresses" or "Accounts using default profile pictures". This can instantly filter out over 80% of temporary bot accounts.

1. Train the Algorithm: Do not engage with spam

In "For You", "anger" is also a high-weight interaction.

• If you click on a marketing account to curse at it, the algorithm will think you are "interested in this type of content" and will push more to you tomorrow.
• Correct approach: Click the three dots in the upper right corner of the tweet and select "I don't like this post", or simply Block. Blocking is the strongest command to train the X algorithm, as it directly cuts off the weight of that account and similar accounts.

Summary Recommendations

If you prioritize information quality, my recommendation is: focus on "Lists", supplement with "Following", and completely ignore "For You". Can X be described as Spamization? Indeed, the term Spamization can be used to describe this phenomenon. It can even be said that X is currently undergoing a systemic collapse caused by both "mechanism loopholes" and "profit motives".

In social media research, this phenomenon is sometimes referred to as the X version of "Enshittification". If previous spam was "hiding in the shadows", then the current X is "mechanistically embracing" spam content.

Here are three underlying characteristics of X's "spamization":

1. Content Degradation due to "Revenue Sharing" (Engagement Farming)

Musk's Ad Revenue Sharing program is the biggest driver of "spamization".

• Logic: As long as there is traffic, verified users can share revenue.
• Result: A large number of "Engagement Farmers" have emerged. They do not produce content, but rather deceive for exposure by copying popular tweets, posting highly controversial traffic-driving posts, or posting AI-generated nonsense in the comment sections of popular big V accounts.
• Consequence: Your comment section is no longer a discussion forum, but has become a "harvesting battlefield" filled with bots and verified marketing accounts.

1. "Blue Check" Priority: A Pass for Spam

Previously, the "blue check" was for identity verification; now, the "blue check" is for weight purchase.

• Logic: As long as you pay, your replies will be prioritized.
• Result: This mechanism completely disregards the "real quality" and "interactive value" of the content. Even if a verified account posts pure spam, it will rank above the quality comments of ordinary deep users.
• Consequence: This "paid queue jumping" mechanism provides institutional protection for low-quality content.

1. Exponential Growth of AI-Generated Noise

With the popularization of generative AI, X has become an experimental ground for AI nonsense.

• Automation: Bots are no longer simple "repeaters"; they automatically generate seemingly reasonable but meaningless comments based on the tweet content.
• Data Pollution: Many "Dead Internet" characteristics have emerged, namely: bots posting, bots replying, bots liking. Real users are pushed to the margins, making it feel like a "digital zombie" carnival.

Evolutionary Landscape of X Spamization

Feature	Traditional Twitter Era	Current X (Spamization)
Filtering Mechanism	Real user forwarding, credit endorsement	Algorithmic weight, paid priority
Main Conflict	Information overload	Spam inversely dilutes quality content
Interaction Purpose	Opinion exchange, seeking truth	Deceiving clicks, earning revenue sharing
Comment Section Experience	In-depth conversations at the top	Verified nonsense/ads at the top

Conclusion

The essence of Spamization is the bankruptcy of social credit. When a platform no longer rewards "depth" and "authenticity", but instead rewards "dwell time" and "paid weight", spamization becomes an inevitable outcome. Topic: For You Interactive Rate Information Noise Follower Chain Driven Content Distribution List Function Spamization Blocking Keywords Underlying Logic Twitter Twitter Blocking Settings Timeline Sorting Bot Filtering Weight Competition Following Water Army Attention Black Hole Traffic Annexation Deep User Social Granularity Algorithm Algorithm Driven Train Algorithm Quality Filter Flame War