The 20x multiplier that drives every social feed

Rest in peace to the reply guy strategy.

X open-sourced the real For You algorithm last week. The advice that built thousands of accounts, “just leave 50 high-quality replies a day”, is now structurally broken.

And it’s not the only thing that died.

The algorithm does not want:

→ curated content

→ low-effort posts

→ unoriginal noise

→ pure reply volume as a primary distribution channel

Even if you barely post on X, this matters.

LinkedIn will never show you their code. YouTube won’t. Meta won’t. This is the only time a major platform has handed over the actual ranking logic. What’s inside explains how every modern feed decides whose work gets seen.

I don’t care about “growing on social” for the cool kid cred. I care about distribution, pipeline, attracting talent, and because social proof definitely gets you connected with wildly talented people.

So I built the X Reach Kit. Free. Open source.

It exists so you can answer two questions: will by post get seen and what change would help me get seen by more people?

It’s open source scroll to the bottom and get it.

A screenshot went viral claiming one report (when someone reports your content) cancels 468 likes.

The math is clean. A report is weighted at -234, a like is weighted at 0.5, and 234 divided by 0.5 is 468. It makes for a terrifying screenshot.

It's also completely wrong.

Think of it like a carnival game.

The carny puts up a giant prize board showing what everything is worth: a giant stuffed bear is worth 500 points, a plastic keychain is worth 1 point. That's the prize list.

What they don't give you is the secret to how weighted the ring toss bottles actually are. That's the odds.

In X's system, a post's score is just those two things multiplied together:

Score ≈ (The Prize Value) × (The Odds You Win)

X published the prize list: a like is worth 0.5, a forwarded link is worth 20.0, a report is negative 234.

What they didn't publish is the brain that calculates the odds. That comes from Phoenix, their private machine learning model that predicts how likely a specific person is to actually take that action.

Without that private AI model, any tool claiming to "grade" your post is just multiplying a public number by a random guess.

That's why no tool on earth can compute a real score for your draft, and it's why my X Reach Kit won’t hand you a fake "99/100" grade.

Every social feed runs on one mechanical problem: there are a billion posts competing for ten slots on a screen. You’re one voice in a crowd of billions.

To solve that, X built a bouncer (and every other social platform does too).

X’s bouncer runs four checks, in this exact order:

When founders tell me their content is flopping, they almost always assume they have a writing problem (Check 2). They rewrite hooks, punch up copy, and buy $4,000/month ghostwriters.

In reality, 50% of them are failing Check 1 or Check 4, where the algorithm drops the post before a single stranger can even see it.

Buried in the code is an uncompromising drop rule: replies and reposts are thrown out of the recommendation pool if the viewer doesn’t already follow you.

Thrown out completely.

The feed will never carry a reply you write out to somebody new. Ever. Same for a repost. The only strangers who see it are the ones who already found that thread on their own.

There is one limited break for small anccounts. Under 1,000 followers, X will grab exactly one of your posts and push it near the top of somebody's feed to give it a shot. Originals only.

That’s why the pure reply guy strategy is finished as a primary growth lever. It was always better as relationship building than as cold distribution. The code just made that crystal clear.

Want to see if your last 20 posts even cleared the stranger gate? That’s one of the things X Reach is built for.

Reply ALGO to this email with your X handle and I'll run your account through the diagnostic tool and send back where you're getting stopped.

Here are the production weights that matter most:

Everything near the top costs the viewer a real action

A like is a twitch of your thumb. So it’s valued low.

But sending a post link to your friend takes effort (and some of your reputation) spends so it sits at the very top.

Every feed on earth is hunting for the stuff people pass along.

So the question stops being "will this get likes + views" and becomes:

is this good enough that one person stops and sends it to one specific other person?

Most creators still optimize for the 0.5 action because likes feel good. The algorithm is optimizing for the 20.0 action.

If you and another account follow each other, a reply from them on your original post carries roughly four times the value in your feed.

X's own commit history shows this boost went live on July 13 and got dialed back a few weeks later because feeds were getting too insular. Nothing changed on Thursday. What changed is that you can finally see how it works.

This isn’t about engagement pods.

Pods are seen by X heuristics. I’m talking about genuine peer networks. Having 20 real operators in your niche who actively hype your work creates an organic distribution floor that’s better than an ad buy.

This is what most creators call a shadowban (and what self-respecting social strategist hasn’t had at least a dozen alt-accounts shadowbanned):

→ Your followers are checked against 28 safety rules.

→ Strangers in the For You feed are checked against 54.

That’s 26 extra roadblocks standing between your post and anyone who doesn’t already follow you (link placement, spam labels, NSFW, AI slop signals, abusive content, etc).

If a post hits with your followers but crashes into a complete wall on reach, it’s probably not your content that’s the problem. You tripped a stranger filter.

Posts older than 48 hours are also mostly filtered from For You. There’s no long tail for most content (unless you’re posting long form video which can resurrect for 30 days, sometimes even 5 years).

A lot of people reading the code claimed Premium accounts get zero ranking boost.

That’s only half true.

And no, there isn’t a raw “pay to go viral” score multiplier inside the algorithm weights, Premium still gets you three critical advantages:

Paying doesn’t make bad content go viral. But in a feed flooded with bots, scrapers and fake accounts, verification gives you a social edge.

A lot of people reading the code claimed Premium gives you zero advantage.

Open anybody's replies right now. You'll see an endless ocean of AI slop instantly. Each post with the same tells. Same rhythm. Same dead eyed cheerleading. Same little ‘beige’ question at the end (and don’t get me started on LinkedIn. It’s 100x worse).

If you get flagged for AI slop, you pick up a label that lasts 30 days, and that label stops you from reaching anybody who doesn’t already follow you.

A full month of organic reach gone for sounding like a template.

Study what works from real people talking. Then say it in your own voice.

The studying, the research. Knowing what works. It’s still the job. It’s why you read good posts and hunt outliers instead of staring at a blank page.

The problem is people treat a template as a finish line instead of a starting line. They never change it. And now that everybody’s AI is running the same three templates, copy-paste is worth nothing.

These social growth tips are dead and buried. If you get this advice from your growth or content people, make like Jordan Peele and get out:

what the x reach kit does

1.  Last 20. Paste a handle.

→ it pulls the last 20 posts and tells you which ones are still open for stranger distribution, which are closed, and which need work.

Shows the mix of originals vs replies/quotes/threads.

2.  Rewrite. Paste a draft.

→ it classifies it as OPEN / CLOSED / REWRITE and gives you THE ONE CHANGE that reopens the closed path (e.g. “Post this as an original, not a reply”).

3.  Under the Hood. Paste X’s Under the Hood readout.

→ it maps the labels to what the published code actually says they do.

4.  Post-mortem

→ For a post that already died or suddenly dropped to followers only. Diagnoses what stopped it based on the real code.

Before you write your next post, ask the two questions:

Is this valuable enough that somebody would send it to a person they know? And who, specifically, do I want them to send it to?

That’s the whole north star. I want to show up in group texts, Slack channels, and DMs.

You might not post on X every day. That's why this matters even more.

LinkedIn will never open-source their code. YouTube won't. Meta definitely won't. You'll spend the rest of your career guessing at those feeds.

Use X as the practice room where you can check your instincts against the answer key, then take the four questions everywhere you publish:

1. Can they hear you?

2. Will they pass it on?

3. Do they know you?

4. Do they trust you?

The platform changes. The order never does.

What you get:

→ Last 20. Paste a handle. Your last twenty posts sorted into STILL OPEN, FILED, CLOSED, and DO BETTER against the published rules.

→ Rewrite. Paste a draft. You get the rewrite, the one change, and paste-ready send / follow / reply lines. No score.

→Under the Hood. Paste X’s own readout. Their words mapped onto published labels, unmatched strings left unmatched instead of guessed at.

→ the shadowban differential: 28 rules for followers, 54 for strangers, and what each extra drop does

→the post-mortem, for a post that already died or an account that suddenly went followers-only

→ 207 machine verified citations, pinned to a commit, with a script that re-checks every one

MIT license. Fork it.

It pairs with Frontrun (finding what your market is talking about) and Slideshow Kit (formatting the creative).

If the repo helps, star it.

If this was useful, do the thing the algorithm actually rewards:

Copy the link and send it to one specific person who is still grinding the old playbook.

Go big.

Matt

P.S. Reply ALGO with your handle and I’ll run your last 20 posts through the tool and show you exactly where distribution is dying.