- The Wizdom Newsletter
- Posts
- X just showed us how to win the algorithm
X just showed us how to win the algorithm
I went through the release. Here’s what founders need to know.

I’ve posted 51,888 pieces of content on X, and my team helps create another 1,923 every month across 60+ founder accounts.
That gives us a pretty unique view whenever something changes.
We have founders who started from zero, founders with hundreds of thousands of followers, SaaS companies, agencies, CEOs, completely different niches.
So when X changes the way content gets ranked or distributed, we usually start seeing the effects across multiple accounts pretty quickly.
And recently X gave us one of the clearest looks we’ve ever had at what’s happening underneath the For You feed.
They released more of the production ranking system, the weights behind predicted actions, the filters that can limit visibility, and a new tool that can show certain accounts which visibility labels have been applied to them.
I spent some time going through everything, and there are a few changes I think founders should actually care about.
Likes matter a lot less than people think
X is predicting what each individual person is likely to do when your post lands in front of them.
The current defaults include:
like: 0.5
repost: 1.0
reply: 5.0
quote: 5.0
DM share: 5.0
copy-link share: 20.0
follow author: 4.0
I wouldn’t look at those numbers and start doing engagement math.
They’re weights applied to predicted behavior for an individual viewer.
What I care about is what they tell us directionally.
A post someone wants to reply to, quote, send to a friend, or follow you because it gives X a much stronger signal about that person’s interest.
That’s a big reason we spend so much time thinking about conversations.
Strong opinions. Useful breakdowns. Specific problems. Ideas people actually want to send to somebody else in their industry.
Those behaviors give the system more information about exactly who should keep seeing you.
Relevance is becoming even more important
X is also predicting negative behavior.
“Not Interested,” mutes, blocks, reports, and posts people immediately scroll past are all part of the model.
This matters a lot for founders because your goal should be getting known by a very specific group of people.
If we’re building the account of a SaaS founder, I want their content consistently landing in front of SaaS founders, investors, operators, customers, and the people already participating in those conversations.
A few thousand of the right people can completely change a business.
And every time those people engage with the content, X gets another signal connecting that founder to the niche.
That’s why we get obsessive about the topics, language, people, problems, and conversations surrounding every account we work on.
You’re constantly teaching the platform where you belong.
X is finally giving us more visibility into account-level issues
This might be my favorite part of the update.
X has separate systems for ranking and visibility.
So there are situations where a post can score well and still be filtered because of labels attached to the post or account.
They’re now piloting a feature called Under the Hood that shows eligible accounts some of those visibility-limiting labels.
The first rollout is going to a randomized group of accounts that are at least one year old and have posted 10+ times in the previous month.
For years, account audits have involved looking at the data and trying to figure out whether something unusual was affecting distribution.
We’re slowly getting access to more of that information directly from X.
If this continues rolling out, it makes diagnosing account problems WAY easier.
Posting more has diminishing returns
There’s also an author-diversity mechanism inside the ranking system.
When multiple posts from the same person are candidates for the same feed response, each additional post can receive a lower score.
I’ve seen people take information like this and immediately turn it into some rule like “only post twice per day.”
I wouldn’t do that.
The bigger lesson is that volume needs a reason behind it.
If you have seven genuinely strong ideas, great.
If you have two great ideas and five pieces you made because the calendar had empty slots, I’d spend the extra time making the two better.
Across our accounts, we care a lot more about the quality of the inventory than hitting some arbitrary posting number.
Your followers are only one part of your distribution now
This is probably the piece founders should understand best.
X is pulling content from the accounts somebody already follows, while systems like Phoenix and SimClusters are also searching outside that network for things they think the viewer will care about.
That’s why I’ve seen relatively small accounts suddenly get a post in front of huge people in their industry.
X is constantly trying to make connections between:
- what you talk about
- who engages with it
- what those people care about
- and who else might care about the same thing.
This is where positioning becomes incredibly important.
For every founder we work with, we define the handful of topics they should own, the problems they should constantly speak about, the people they want to be associated with, and the conversations they need to show up in.
Over time, those signals start stacking.
And once X understands where you belong, discovery gets a lot easier.
What I’m actually changing because of this
Very little about our underlying strategy.
We already review every account weekly.
We look at which topics are winning, which formats are working, where qualified engagement is coming from, which conversations are moving inside the niche, and what we should test next.
Then we build another week from that information.
That process keeps us from becoming dependent on whatever algorithm hack is getting passed around that month.
X is going to keep changing.
They’re publishing more of the production configuration now, and they plan to keep exposing meaningful experiments as the system evolves.
So my job is pretty simple.
Keep watching.
Keep testing.
Figure out what actually changes performance across the 60+ accounts we have access to.
And tell you when something is worth caring about.
You have a company to run.
You probably don’t need to spend your Thursday afternoon reading X ranking code.
I’m happy to do that part.
Some Things For You:
We build and run personal brands for founders across X and LinkedIn. You give us around 45 minutes a week.
Our 12-week done-with-you program where we install the strategy, systems, AI workflows, and frameworks into your business.
My full breakdown of how we’re using AI to build personal brands right now.
— Wiz