I spent 2 days inside Twitter’s brain

I did something stupid last week.

I downloaded Twitter’s open-source repo that decides what shows up in your feed and read it.

Line by line.

A few days, multiple shots of espresso, one very supportive wife later.

I found that the “algorithm” is basically a giant calculator trying to predict which posts will keep people scrolling.

A scoreboard. Here’s how it works in plain English:

Every candidate post is assigned points. Highest scores bubble up. Lowest scores die in the void.

What gets you points?

> Dwell time: If people sit with your post, the score jumps. Thirty seconds of reading beats a hundred mindless likes.

> Replies: Conversations carry more weight than likes or even retweets. One thoughtful comment outweighs everything else.

> Novelty: The algo literally compares patterns. If your content looks like recycled templates, it gets marked down.

> Graph fit: Are you talking about what your followers care about? If you drift too far from your lane, you get dinged.

> Healthy engagement: If people reply in good faith and others join in, that’s a positive multiplier.

If it looks spammy, you get downranked.

Now, what kills your score?

> Links in the first line. (The code flags this as a low-quality signal.)

> Hashtag soup. Anything over two looks like spam.

> ALL-CAPS rants. Yes, the code penalizes shouting and I am guilty of this a couple times on the TL.

This means for your strategy:

1. It doesn’t care about follower count. A 5K account with engaged replies can outperform a 50K account posting generic value.

2. It doesn’t care about how many times you post. Frequency without depth just teaches the algo to ignore you.

3. It doesn’t care about looking polished. Perfect graphics don’t move the needle. Depth, specificity, and novelty do.

The algo is agnostic to ego. It only cares about behavior and the only question it asks is:

Does your post keep people engaged and talking? 

If you’re a founder, exec, or agency owner, this what this means in practice:

> Engineer dwell time. Don’t rush your reader. Use story beats, bullets, reveals. Make them sit with you for 30+ seconds.

> Seed conversations. Ask questions or challenge someone’s patterns to spark replies.

> Stay novel. Tie posts to your own data, client work, or lived experience to get original angles.

> Own your niche, If you’re DTC, talk DTC. If you’re SaaS, talk SaaS. Don’t chase broad “business” takes. The algo amplifies authority in your lane.

> Avoid penalties

After two days in the code, the big revelation is… it’s not magic.

It’s a glorified engagement accountant. Add points for depth, novelty, and conversation. Subtract points for spam and shortcuts.

That’s it.

And yet most people will still build their personal brands like it’s 2022 by aiming for virality, posting generic threads, joining spammy engagement pods.

You’re fighting the algo instead of feeding it.

We spent two days tearing through this code because we want our clients’ accounts positioned perfectly to thrive as the platform evolves.

That’s why their reach doesn’t collapse when the rules change. It’s why they’re always in the right conversations on X and LinkedIn. And it’s why they keep closing deals while everyone else is panicking about engagement drops.

If you want that kind of unfair advantage → Book a call with Mogul Media 

Talk soon,

– Wiz

P.S. The irony? By writing this email, I’ve probably tripped the “long dwell time” filter. If you’ve read this far and if we were on X, the algo would already be rewarding me 😉