Understanding social media algorithms to get more advantage + agency (longer read)

In a break from the usual building updates for @Meet-Ting, I’ve been trying to find the time to write this piece up as I care about it a lot. It feels important - and timely - and valuable for the PH community.

Two weeks ago I saw Yuval Noah Harari at the Barbican. I love his work, and as I’ve got older I’ve become more “learn from history.”

Human behaviour seems to repeat - because, at least in my non-scientific understanding, the only thing that doesn’t change is that we’re human.

His talk (for Nexus) was about social media algorithms - and now AI - reshaping us. It was brilliant and a little sad.

After a career working inside these platforms and algorithms, and now building with LLMs, I left conflicted.

The debate that stuck with me: agency.

Algorithms quietly run most of our lives. Finance went from a bell ringing at 9am to a 24/7 casino of algorithmic trading.

But Yuval’s focus was media - how algorithms decide what we see, and of course the good and bad that it brings us all.

Most people treat “the algorithm” like a black box. But if you understand the job it’s doing, you can get some control back.

And more importantly for this post - in our case as builders - it also means we can get an advantage for our product by getting attention.

What’s an algorithm?

A step-by-step way to solve a problem. Which means every algorithm has a job:

• Example (TikTok): the job is matchmaking - pairing a video with someone who’ll care. Success = full completions + replays. Translation: design for finish-ability and rewatch-ability.

• Example (Instagram): there isn’t one master algorithm. Feed, Explore, and Reels each have different goals. Reels cares about watch time and share-ability, so you need to think about % of view completion and getting shared in DMs.

This is why so much “brain rot” dominates our feeds - it’s engineered to be watchable, repeatable, shareable...

Cold-start, niches, and “gas station fights”

Algorithms need to know you. When they don’t, they default to mainstream/viral. That’s why your first week on a platform feels off.

Back in the earlier days of TikTok, this was our #1 challenge - how to make a new users first few sessions more valuable and reduce churn. In the absence of time with the user, the default is popular content on the platform, which at the time was dancing & lip sync, not ideal for everyone and it didn't help for platform perception...

X's head of product tweeted this last week: https://x.com/nikitabier/status/1969057240578748723 = same story, different day.

Over time, tiny signals - what you watch, skip, or linger on - nudge it closer to your real interests.

You’ll still see “gas station fight” clips. Why? You interacted with it, your account is thin on signals, or it’s so viral it breaks out of niche into mass.

The trade-off: better personal relevance, less shared pop culture. Fewer monoculture moments, more personalised micro-cultures. I do miss the days of Harlem Shake and global internet moments.

Platform priorities

Platforms also boost formats they want to grow to stimulate more adoption.

If short video is strategic, Reels/Shorts get oxygen (I can tell you from watching them roll out from the other side they were PUMPED).

Same as the cycles of medium video to long video on YouTube. Why? You can serve more ads in long video.

Know the priority each quarter per platform, and you know what gets extra lift.

Your agency: shape what you see (and what sees you)

If you’re a user:

Train it intentionally. Engage with what you want more of, swipe past what you don’t.

Use tools: Not Interested, keyword filters, unfollow/mute.

Cold-start on a mission: only engage with your ideal niche.

If you’re a founder/startup/creator:

Design for the success signal (completion, shares, replays).

Version for the room (Feed ≠ Explore ≠ Reels).

Ride new formats early - they get bonus reach.

Cluster by theme so the system can recommend you faster (avoid generic topics, design by category/user problem).

Why write this?

People talk about “the algorithm” like it’s magic. It’s not. It’s just a machine with a job to do. Same as AI, really.

Understand the job, and you get agency back - whether that’s protecting your attention, or finding growth as a brand, creator, or startup.

I love connecting people, it's been my entire career and now the mission with Meet-Ting. If you’ve read this far, thank the algorithm for bringing you here - and thank you for choosing to spend your attention with me, when you could have spent it anywhere else!

Dan