The new bakery by the station was almost sold out by early afternoon. Going first thing tomorrow
Hide what you don't wantin your own words
Hide posts you don't want to see with a sentence you write yourself. Works on your X timeline and notifications, in YouTube comments, and on Threads
Filtering accuracy for posts in other languages has been tested less extensively than for Japanese.
What's left isworth reading
Only posts that match the sentences you wrote are hidden
Describe unwanted posts in a sentence
Posts are judged by meaning, not keywords, so rephrased posts are caught too. Sarcasm and indirect wording can slip through
- Posts that look down on and mock people who are trying hard
- Posts that insult someone's looks or character
- Posts that lecture or criticize when nobody asked
- Posts that reveal what happens in the latest episode of [title]
- Posts promoting products with links
- Bait posts like “90% of people get this wrong” or “answer in the replies”
- Soccer results for teams I don't support
- Posts that say “that's just your opinion though” to win an argument
- Topics where someone is being piled on
- Posts about [game title]
- Posts promising easy money and steering people to DMs or a profile link
- Posts that flex salary or education to pull rank
- Posts that put someone down by comparing them to other people
- Celebrity gossip
- Depressing crime news
- Add to Chrome and write your own
Who it's for
For people who post
If you run more than one X account, link a rule set to each one and it switches automatically to match the account you're viewing. Hide the attacks in your replies and spend your time on opinions and questions
Comments that attack the person instead of the content
For people worn out by scrolling
Hide pile-ons, grim news, one-upmanship, and clickbait one sentence at a time, and get a timeline you can look at calmly
Topics where someone is being piled on
For people avoiding spoilers
Write "Posts that reveal what happens in the latest episode of ○○" to hide posts about the plot or the result. Show them whenever you want
Posts that reveal what happens in the latest episode of ○○
For people dealing with harassment
Hide aggressive replies and mockery aimed at you and keep your distance. Unlike blocking, nothing happens on their side
Aggressive replies aimed at me
Focus on repliesthat matter
Hide hostile replies to your posts and comments on your videos with sentences you write. Spend your time on the opinions and questions worth reading
On the paid plan, the same rules also apply to DMs from people you don't follow (message requests). It stays off until you turn it on and never reads conversations in your inbox
Hidden, not deleted
Filtering that keeps up
- Usual wait for a judgment
- 0.4 s
- Almost every post (95%)
- Within 0.6 s
- Time our fast, purpose-built AI model takes to check one post
- About 0.14 s
Measured by sending made-up posts from Japan to the filtering servers. Times vary with your connection.
Lunch at the diner near the office. The daily fried chicken was huge
The new bakery by the station was almost sold out by early afternoon. Going first thing tomorrow
Hidden because of “Posts that look down on and mock people who are trying hard” ·
My kid rode a bike without training wheels today. One full lap of the park, very proud
The forecast said rain, but it's sunny. Hung the laundry outside
Three months of morning runs, and today I finally ran the full 5 km. Feels great
Stays out of your way
Keeps social media fast
On a test page shaped like X, it used about 3.1% of one CPU core while scrolling continuously
No internal systems
It only reads the posts the site has loaded. It never calls the internal systems of X, YouTube, or Threads, never acts for you, and never asks for your password
Ads stay put
Ads on X aren't checked and are shown just as advertisers intended. Only how your own screen looks changes; your posts and follows are untouched
The CPU figure compares a test page shaped like X with and without the extension. It varies by device.
Post text and rules are not used to train AI
Sent
- Text of loaded posts
- Your rules
- An ID for counting daily checks (your account when signed in)Stops at OneFilter
Not sent
- Social media account names
- Browsing history
- URLs
OneFilter's check
- Doesn't store post text
- Keeps it out of logs too
The AI that checks
- Receives only text and rules (for short replies, the text being replied to too)
- Not used for training
- Not stored (except with your own key)
With your own key, the extension sends straight to the AI that checks, skipping OneFilter
Only the verdict comes back: hide or show
Post text is not stored on OneFilter's servers, and the filtering provider is not allowed to store it either (with your own key, TypeSafe AI's terms apply)
Privacy policyEarly access
Get an invitation and join the paid plan during early access, and your first month is free
- Without signing inUp to 300 checks per dayFree
- Signed inUp to 500 checks per dayFree
- Paid planUp to 3,000 checks per day¥300/month