Shadowban Scanner has been updated to v4.0.3

Shadowban Scanner has been updated to v4.0.3!

Here are the main changes in this release. You can view the full list of changes on GitHub’s release page.

Main changes

The main changes in Shadowban Scanner v4.0.3 are as follows. This release mainly addresses changes to X specifications, and there are no new features.

• Fixed an issue where messages were no longer displayed on posts due to changes in X specifications (Robot-Inventor/twi-ext#436)
• Updated dependencies

X-related news

Although not directly related to the Shadowban Scanner update, here is some news related to X.

X published the algorithm for the recommended timeline on January 20, 2026.

In relation to this algorithm, there have been posts claiming things like "a like on a post multiplies the score by X", "a repost is X times", or "a reply is X times a like", but please note that all of these are misinformation.

It is true that likes, reposts, replies, and other actions are not all treated with the same level of impact, and that coefficients are applied to adjust their impact (weighting), but the specific numerical values have not been made public.

*Added on February 5, 2026: Strictly speaking, in the latest algorithm, recommended posts are selected not based on the number of likes or reposts themselves, but on a score by weighted predictions of the probability that users will perform those actions.

In the published source code, there is a prominent comment regarding parameters such as weighting that says "Excluded from open source release for security reasons."

In other words, while it is true that scores are adjusted for each type of action such as likes and reposts, any post that mentions specific numerical values is misinformation (or would be leaking internal confidential information).

In older source code that was published in 2023, when X was still Twitter, numerical values were also made public, but those values were later removed from the publicly available source code.

As mentioned above, the fact that specific numerical values are not public is clearly stated in a prominent place in the source code, so posts that mention specific numbers are essentially written without reading the source code.

They were probably copy-pasted from AI answers without verifying their accuracy.

When you ask AI about X's algorithm, AI may refer to outdated information on the internet or fabricate inaccurate information (hallucinations).

Even if you attach the URL of the published algorithm and ask AI about it, AI often answers without reading the source code, so caution is required.

Reading the source code yourself is the most accurate approach, but if that is difficult, it is better to attach the algorithm's Zip file and have AI analyze it, or use a feature that connects AI to a GitHub repository.

In any case, the posts that appear every time an algorithm is published, claiming things like "a like on a post multiplies the score by X", "a repost is X times", or "a reply is X times a like", are misinformation.

When X's algorithm is published, and when the terms of service or privacy policy are updated, many related pieces of misinformation tend to be posted.

Please be careful not to be misled by such misinformation.

Request for support

Shadowban Scanner is a hobby project by the developer, but it takes a lot of time to develop and maintain. If you have the means, I would appreciate your support for its development.

I accept donations through Buy Me a Coffee or GitHub Sponsors. Please note that there are options for both recurring and one-time donations.

Also, simply posting reviews or ratings in the store, or following the author's X (Twitter) account, helps motivate development. If you are knowledgeable in programming or interested in contributing to translations, please consider joining the development on GitHub!