X finally open-sourced its algorithm: Here are real tips to grow your reach in 2026

X dropped the code for its new recommendation algorithm on GitHub this week, and people wasted no time tearing it apart to figure out what actually gets you seen on the platform.

After years of complaints about shadowbanning and mysterious reach drops, the company finally let everyone peek under the hood at how the For You feed decides what shows up.

The new system runs on a Grok-based transformer model, the same architecture powering xAI’s chatbot. What makes this different from the 2023 open-source release is that X eliminated almost all hand-coded features. The algorithm now learns purely from engagement patterns, which means old tricks like hashtag stuffing and posting at “optimal times” won’t do much anymore.

Developer @vibhu published a detailed breakdown on X, reverse-engineering the entire ranking pipeline. According to his analysis, the system pulls posts from two sources: accounts you follow (called Thunder) and random posts from across the platform (Phoenix Retrieval).

Each post then gets scored based on predicted actions like replies, likes, reposts, profile clicks, and even how long you’ll spend reading it. These probabilities get multiplied by hidden weights and added up to create your final ranking score.

Here’s where it gets interesting. The weights matter way more than most people think. Research into the code shows that replies carry a weight of 13.5, while likes only get 0.5. If someone replies and the original poster responds back, that interaction jumps to a weight of 75. Profile clicks score even higher at 12.0. Translation: one reply where you engage back equals roughly 150 likes in terms of algorithmic value.

Several users on X have been sharing their findings. @tvytlx noted that videos and images get direct boosts, early replies fight time decay better, and links in your main post can kill dwell time.

@Sarabjeet___ pointed out that bookmarks act as huge multipliers, and replying to your own posts early helps build momentum. Others like @iamspacecreated and @3szhAl started threads on consistency and viral tactics.

The flip side? Negative signals tank your reach fast. Blocks, mutes, and “not interested” clicks all carry negative weights that push your content down. If your post gets shown to someone who then blocks you, that’s particularly bad for future distribution.

What this means in practice: you want posts that spark conversation, not just agreement. Leave hooks that make people want to reply. Ask questions without spelling out the answer. Be slightly contrarian without crossing into spam territory. Use images or videos when they fit. Build a clear niche so the embedding system knows what kind of audience to show your stuff to.

Even Elon Musk admitted the algorithm needs work, calling it “dumb” while promising ongoing improvements.

X plans to update the open-source code every four weeks, so what works now might shift. But the core insight probably won’t change: the system rewards genuine engagement from people who actually want to see your content, not passive scrolling.

We stand out from the tech-media crowd because we break news stories; we mainly bring you stuff that you won’t find anywhere in the mainstream tech media. Our stories have been picked up by some of the world’s most popular websites and media outlets—more info is available here.