It’s been interesting since early July, with Threads and Twitter going head-to-head. They’re on opposite tracks. Twitter, a text-based social network with approximately 250 million daily active users, began throttling usage, limiting users to a few hundred tweet impressions and then blocking them for a day. Instagram Threads, on the other hand, has been growing its user base with record numbers of subscribers since its inception. In Korea, the pronunciation of ‘thread’ has led to a variety of buzzwords using the word ‘tsu’. Examples include ‘Tsipal’ and ‘Tsiseng’.
Twitter streamlines systems amid usage restrictions
Elon Musk, owner of Twitter, shares the following. Roughly speaking, it’s about limiting the number of reads and posts per day to prevent system abuse.
The restriction remained in place for the next 2-3 days. When using one of our paid Twitter accounts, Tweets were not loading. Even after the restrictions were quietly lifted, other restrictions were imposed, such as logged-out users no longer being able to see Tweets. This means that, at least for now, you can’t use Twitter for public announcements that anyone can read.
The exact cause of these actions is still unknown, but it’s clear that they are related to infrastructure issues. Twitter operates its infrastructure in three of its own data centers. Twitter is leveraging Google Cloud to some extent, and based on multiple sources, for machine learning (ML) use cases as well
Is malicious behavior detection just an excuse?
Why did you have to throttle your network (force performance down to prevent system overload)? A few days later, Twitter’s official explanation was very vague. Twitter said in a statement that it temporarily throttled usage “to detect and remove bots and other bad actors who are harming the platform.”
Bots and bad actors exist on every social media platform, and detecting, restricting, or banning these accounts is typically done by machine learning systems that detect suspicious patterns. How does Twitter perform this detection? According to an engineer we spoke with, Twitter relies on Google Cloud’s ML and AI services. And the day after my contract with Google Cloud expired, I was throttled.
On the infrastructure side, Twitter currently operates three of its own data centers, with plans to reduce this to two. However, Twitter engineers I’ve spoken to in the past have said they use Google Cloud because it’s more efficient than doing it in their own data centers.
A month ago, a match called Platformer reported. Source link
“Before Musk bought Twitter last year, Twitter had a multi-year agreement with Google to host services related to fighting spam, removing child sexual abuse material, account protection, and more. (…)
Twitter kept demanding that Google lower its cloud costs, at some point Google stopped paying its cloud bill, and now it’s planning to leave the platform altogether.
The move has sparked a movement within Twitter to move as many services off Google Cloud as possible before the contract ends on June 30. But the project is running behind schedule and leaving some questions unanswered, sources said.”
A few weeks later, on June 21, Bloomberg reported that Twitter had resumed payments to Google Cloud and mended their relationship. But I wonder if Twitter simply negotiated a much lower Google Cloud quota starting July 1. Or would they lose the tools to reliably detect spam accounts and bots and be forced to limit their platform because they needed time to build their own approach?
Mitigate the problem by limiting Twitter users
Either way, Twitter decided to mitigate the problem by reducing engagement on the platform and limiting users. Since fewer impressions overall means fewer ad impressions, it’s likely that this move has resulted in some reduction in Twitter’s ad revenue. It also highlights the difficulty of running a large social media network reliably when cost-cutting is a top priority, as Twitter has been for months.
In contrast, Instagram threads were launched in a big way with no throttling. Threads is sure to be a challenger to Twitter by offering the same text-based social network features. The app is off to a great start, reaching 10 million subscribers within seven hours of launch.
Threads, a long-established service that wins the day
The pace didn’t slow down much after that, and in the first 24 hours, the thread surpassed 30 million subscribers. Onboarding 30 million new users in less than a day and not imposing rate limits is a huge accomplishment for any social network.
For example, another Twitter challenger, Bluesky, launched its app in February 2023. In the four months since, the app has grown to 180,000 users, but 1.9 million are on the waiting list. Bluesky is throttling user onboarding because it needs to scale its infrastructure as the app grows and build out moderation tools like blocking or muting, which are becoming increasingly important as the social network grows.
Meta’s ability to launch without throttling is very impressive at a scale of tens of millions of users for two reasons.
- Infrastructure ready from day one: Thread had the infrastructure in place – compute, storage, networking, etc. – to quickly scale to millions of active users. Being able to leverage Instagram’s infrastructure has been a huge benefit to the Threads team. Instagram has over 1 billion monthly active users.
- Monitoring tools and network dynamics: Monitoring is critical to growing a social network. This includes both self-management, where users have tools to block or mute other users, and platform-level management, such as removing hate speech or addressing harassment. Once again, this is an area where Meta could really benefit from the expertise she’s gained on Facebook and Instagram. Of course, the Threads team relies on it too.
The threaded launch we’ve seen so far is perfect, except for one detail that isn’t. The app is available worldwide, but not in some European countries (EU). This limitation may be related to the fact that the app is not yet GDPR compliant. For example, to delete a threaded profile, a user must also delete their Instagram account.
If there’s any doubt that threads will take Twitter’s place, Mark Zuckerberg, CEO of Meta, has an answer.
The next few months will be interesting, considering that Thread is clearly pursuing a growth strategy and is not worried about infrastructure costs. Meanwhile, Twitter is optimizing its infrastructure costs, even at the expense of declining engagement and subsequent growth. So who will be the winner?
