🐫 The eye of the needle

How should we make sense of Threads?

Threads is live! Please find me here.

It will probably be the fastest-growing consumer service to 100m users in history, faster even than ChatGPT, securing 30m signups in just two days.

Twitter is broken. I’ve used it for 16 years, I built a company on it. And for years I used it as a source of professional information, often supplemented by third-party tools like Prismatic, RightRelevance and Nuzzle. In its early years (till about 2013), Twitter really suffered from noise pollution: scams, tweet circles, retweet to win, and the rest. It was one reason why PeerIndex existed, and why, in 2011, we launched a bot-or-not detector for Twitter.

Twitter’s investment in trust, safety and community health dealt with much of that. What it couldn’t handle was the culture war, Trump and Brexit. In the period after 2016, the social and algorithmic filters I had set up over years become largely useless.

The second breaking of Twitter has been Musk. I’m Justice Potter Stewart in evaluating why Twitter is broken and getting worse.

But it is definitely worse for me. Every interaction yields less interesting material. Specialist discussion tapers off more quickly. The ‘For You’ feed is a crapshoot. My lists, curated over the years, are lower quality, lower velocity than in ages. True, AI discussion is significant (although polarised) but many of the other topics I followed (and people too) have disappeared or become full of verbiage, grandstanders and trolls.

We haven’t seen the bottom yet. I’m not convinced that Musk can improve Twitter. But we shouldn’t count him out. And for all its weaknesses, it doesn’t really have competitors as a news backplane. So I’m still using it regularly and have become unseemingly generous with the block button.

A new thread

But perhaps Threads might be it.

The complexity for any social network is, of course, the network. Migrating 16 years of a social graph¹ is probably impossible. (Which is why I’d like you to follow me on Threads).

Just a screenshot from my Threads app

Starting from scratch, as Mastodon and BlueSky, have done will be tricky. Ultimately, the move to Mastodon or BlueSky worsens your user experience. With the exception of a few communities, there aren’t many people there. And my reason for using Twitter (and, by extension, any competitor) is to understand what people (including practitioners and experts) in specific domains are thinking. I’m boring. Not really there for the lulz, more for the learning.

I still get more material from Twitter than I do from those platforms.

The reasons to move to BlueSky or Masto’ are not so much about short-term needs. They are long-term: that decentralised social networks are more self-sovereign than centralised ones.² That over the years, it’ll be better if we supported federated social media more than mercurially-owned corporate versions.

But the immediate experience is worse. I can’t find all the experts or communities of practice I used to follow on BlueSky and Mastodon. More complex apps and tricky sign-on flows don’t help.

Could Threads address this? It will bootstrap from the Instagram network. This means potentially hundreds of millions of people easily trying the app, including many who have not made their way with Twitter. Threads could quickly have more monthly active users than Twitter.

But… that doesn’t mean Threads will replace the Twitter of old, the Twitter of smart conversation (yes, in a sea of other material).

The problem is that people don’t have unitary social graphs. My Twitter graph is very different to my Facebook, Instagram and LinkedIn ones. We knew of this heterogeneity from the very early days of PeerIndex (we cross-indexed multiple social networks and could evaluate the differences in graphs and graph structure by app.)

Not only are graphs different but people’s behaviours are different. My Instagram feed only has about 4,000 followers (many of whom think I am a villainous minor Bollywood actor), compared to a Twitter following of 43,000 and LinkedIn of more than half a million. Each one of those graphs exhibits very different behaviours. I follow a couple of thousand people on Twitter and track some 10-15 lists (of about 100 users each). The cost of moving that over to Threads will be extremely high.

Since the graph is the collective sensing network that delivers the material who is in the graph and how they behave will determine your user experience, recommendation algorithms notwithstanding

Brute force

Threads is using a brute force bootstrap. It’ll pour your Instagram network into your Threads one.

This might overcome the cold start problem. But it isn’t clear it solves to community start problem. Our social graphs reflect particular communities of practice or relationships in which we participate. Put it this way: you might love a big family gathering at Christmas… when it’s your family. But teleported into someone’s random family gathering is just a bit weird.

For Threads to become a realistic competitor (in terms of time and attention), it’ll need to help us with that graph problem. Yes, product features, like lists and hashtags, will be essential. But ultimately, social networks are about networks. Curiously for non-Twitter users, those who struggled to build the right graph on Twitter, Threads might appeal… for someone like me, where there is minimal congruency between my Instagram and Twitter graphs, it is more of an issue.

The fresh air of competition

But the most exciting thing about Threads is that it’s actually a genuine bit of entrepreneurialism from Facebook. Look at Zuckerberg’s double pivot this year. The metaverse - tens of billions of sunk costs for legless avatars and empty virtual worlds - dumped for a constant flow of AI projects and now a scrappy, yet polished, new app.

Yes, even with all the privacy issues (to which we’ll refer in Sunday’s newsletter) and problems that Meta products bring, competition is quite healthy. It may prompt better behaviour from Twitter. It might even encourage other firms to figure out how they might nibble at social-network-infused products. Apple would be an obvious one: it can tie users back to real verified IDs, it has a business model that doesn’t depend on attention or advertising, and Cupertino often sees its products in social contexts. The firm has failed twice before with community apps:eWorld and Ping were both lead balloons. But little-by-little social network capabilities are appearing on their platforms.

In the small part of the world that is consumer internet apps, that is pretty exciting. Big successful companies are not renowned for trying things like this. So I’m happy that Meta did.

It is after all harder for an incumbent to innovate than it is for a camel to pass through the eye of a needle.

Cheers,

Azeem

1. I use “network” and “graph” interchangeably here because of this.

2. But decentralisation isn’t all it’s cracked up to be. Like it or not, centralised functions (around moderation, verification, or safety) are extremely valuable. It’s one reason many people prefer Apple products to Android ones. Even in the case of Twitter, the new management team realises they need all of those capabilities. The issue was not whether they believed in centralised functions of platforms, it was that they just didn’t like the way they were run by the previous management.