Giving X users real control over their feed
People don't control their X feed, they cope with it: mass muting, alt accounts, or quietly leaving. Real control over the algorithm is buried, blunt, or missing.
I designed the missing layer inside X's own patterns, one lightweight and one granular: feed modes and an interest dial that make customizing the feed easier and more personal. Four of five research participants came back to test it and finished every task.
A feed I waited years for
Do you know where to customize your X feed? Most people never find the controls, they just cope: mass muting, alt accounts, or deleting the app.
Fun fact: I didn't have an X account until I came to the States in 2021. Not mainly the Great Firewall (a bit, yes), but the app overwhelmed me: walls of text I had to decode, as a non-native English speaker, before I could decide if I cared.
Breaking into tech pulled me back. The same wall was still standing. You either take what the algorithm serves, or mute your way to sanity.
All algorithm, no agency
People don't control their X feed, they cope with it. Five interviews showed what stands between coping and control.
The algorithm runs the show
X decides what you see, and the handful of controls that do exist barely move the needle.
Interests pile up and blur together
Topics you cared about months ago keep resurfacing, and work and play share one feed, so people split into alt accounts to keep them apart.
Controls hide where no one looks
Real control sits in settings nobody opens, far from the moment of irritation.
“If I have to go into a settings page and manually configure it, I'm way too lazy to do that.”
from the interviews“I'm on X for laughs and giggles. If I see Political or BBC, why do I care?”
from the interviewsPeople want granular control over their feed.
The interviews were full of people taking control the destructive way instead: a second account, a blanket mute, or deleting the app.
The tools that exist are hidden, ineffective, or nonexistent.
What X ships is buried in settings nobody opens, and blunt once you find it, so mute wins by default.
Those needs didn't point to one average user. They split into two mindsets, and any solution had to serve both without forcing friction on either:
How might we give people meaningful control over their feed without turning it into a chore, so users across different needs get a feed that actually fits them?
Two surfaces, one system
The research set the bar: controls that
Live inside existing workflows
Offer immediate utility over explanation
Separate contexts without alt accounts
Are findable at the exact moment of need
Two features clear it together, one lightweight and one granular.
Feed modes that live within thumb reach
Monday morning, you open X for the industry, not for football. One tap on the menu and the feed follows. One mode runs at a time, so the contexts never blend.
An interest dial instead of a blunt mute
The election ends, or you finally catch up on the show. Instead of muting a topic forever, you turn it down: less Celebrity News, for 30 days, right from the post.
Two separate controls would have rebuilt the problem
My first instinct was a mode switcher for Explorers and an interest tuner for Intentionals. Separate controls meant separate places to find them, which is the scattered-settings problem people already ignore. Both moved into one connected experience, and every quick tweak writes itself into the main settings.
Where the workarounds came from
Everything above traces back to five interviews about how people actually manage their feeds. Two questions went in with me:
Why workarounds beat the controls X already ships.
What else they need that the feed doesn't give them.
The choice isn't control versus engagement. It's designed control versus improvised exits.
The honest objection: giving people control could reduce the engagement the algorithm is built to maximize. But the interviews showed people already taking control the destructive way, splitting into alt accounts, muting entire topics, or, like me for years, just leaving. That churn is invisible to engagement metrics.
From insights to wireframes
The features took shape in low fidelity first, testing where control should live before polishing how it looks.
Three features, drawn more than once
Low fidelity settled where control should live. Mid fidelity settled what each control actually is, and all three earned their final shape by losing an earlier one.
Should custom modes sit beside X's own tabs, or above them?
Option A listed all four as equals; option B split them into two tiers. The shipped card takes B's structure and keeps A's rule that only one mode runs at a time: Following and For You describe where posts come from, Work and Personal describe what they're about.
How much precision can one topic carry before tuning it becomes work?
Two drafts missed: a slider mixed with a dropdown, then five named stops, both taking over the screen. It shipped as one gesture on a card that stays on the post you're reacting to, with three stops instead of five.
Once a mode exists, where do you go to fix what's in it?
The first pass managed a mode by deletion, a trash icon on every tag. In the final pass every tag is a chip you can enter, and topics tuned from the feed land in their own Adjusted Topics group.
Three explorations, one answer
Every fork resolved the same way: give the control a place in a hierarchy instead of another switch in a list.
Where it landed
The finished feature, surface by surface.
Feed modes
Feed modes live one tap from the menu icon.
The mode card is where you tune what each mode means.
The interest dial
The interest dial sits on the post you're reacting to.
Every topic you've touched collects on one managing page.
Every task completed; two frictions became the roadmap
This is a concept validated in testing, not a shipped feature. The roadmap comes straight from what testing surfaced: move the customization entry up the hierarchy, and add a first-use explainer so a mode switch earns trust before the first tap.
In February 2026, a year after this concept, Threads shipped Dear Algo: post a request like 'show me more posts about podcasts' and your feed adjusts for three days. Platforms are arriving at the same bet this study makes.
Lessons learned
The average user is a fiction
Designing for the middle of the spectrum would have hidden exactly the mindset differences that drove behavior. The two ends were the real brief, and keeping every new control optional means people who never want to tune anything never have to meet them.
Systems thinking for a mature ecosystem
Going in I didn't have a systems-thinking lens, so I spent a lot of time working out how to fit new solutions into what X already had. I got faster at asking where X already solves this before drawing anything new. Next time I'll ask it on day one.
That's the end of this story. If you're still curious, here are two more.