Little monsters, algorithm shifts, and the metagame
Please, take my weekly links – it's all I have
Welcome to this week’s free edition of The Terminal. If you’d like to become a subscriber and gain access to subscriber-only posts and the archive, hit the button below.
Facebook employees were recently given a new directive with sweeping implications: make the app’s feed more like TikTok.
Simply bringing Reels, the company’s short-form video feature, from Instagram into Facebook wasn’t going to cut it. Executives were closely tracking TikTok’s moves and had grown worried that they weren’t doing enough to compete. In conversations with CEO Mark Zuckerberg earlier this year, they decided that Facebook needed to rethink the feed entirely.
In an internal memo from late April obtained by The Verge, the Meta executive in charge of Facebook, Tom Alison, spelled out the plan: rather than prioritize posts from accounts people follow, Facebook’s main feed will, like TikTok, start heavily recommending posts regardless of where they come from. And years after Messenger and Facebook split up as separate apps, the two will be brought back together, mimicking TikTok’s messaging functionality.
On the face of it, this is pretty unremarkable. Obviously Meta is going to want to resuscitate its platforms as they bleed users to TikTok, and the clearest way to do that is to clone what makes TikTok so successful.
But it’s also interesting to think about the converging feature sets of platforms as a reflective of broader social trends. Facebook’s original mission, quite famously, was to replicate your real-life social relationships in the digital realm. You’d add people you knew, look at their photos and read their various life updates. TikTok, on the other hand, operates on a completely different model. It often doesn’t seem to care very much who you’ve elected to follow, and will gladly never show you any of their content on your For You page if its inscrutable algorithm decides it knows better than you do.
Back in 2018, and amid an unstable political environment which blamed Facebook for basically destroying civil society, the company announced it was going to deemphasise engagement-rich politics and news content and get back to its core focus of displaying what your friends and family share. Obviously, this didn’t take the heat off the company, and it hasn’t made anyone more excited to use Facebook. (They simply can’t escape politics either, apparently: new research argues the algorithm change benefited Republican groups more than Democratic groups in the US.)
It seems the original Facebook mission of connecting you with people within your tangible social circles just isn’t very compatible with the kind of social order that Meta’s new target audience enjoys. I feel like a lot of the Covid-era discourse around alienation and parasociality emerges in part in this context; a world in which we’re more comfortable with allowing the algorithm to not only show us relevant content but define the actual contours of our online relationships.
Who knows how far Meta actually goes with this. The Verge story contains some placatory remarks from executives who insist that content shared by your close network will always be central to what they do. But I think it’s actually more interesting than just another algorithm change by a wheezing platform.
Crash and earn
Company executives are trying to give the impression that nothing is wrong, but a clear sense of tension has edged in since Axie token values began to plummet late last year. When I first spoke to Sky Mavis co-founder Jeffrey Zirlin in late January, he told me he was living somewhere in the US but paused when I asked where in the country he was. “I could live anywhere, I don’t usually leave my room,” he said. He did finally give me a more specific location, but asked me not to make it public, noting that his team has gotten death threats. “We have to be careful revealing our location, just like the president doesn’t always have to reveal his location,” he said. “We’re kind of like heads of state.”
Zirlin said he empathized with people who’d lost money—life-changing sums, in some instances. But he added that a crash that got rid of Axie profiteers could have its upside, too. “Sometimes having to flush out the people who are just in it for the money,” he said, “that’s just the system self-correcting.”
An amusing thread about Rainbolt, the leading light of a cool new subgenre of guy you simply must become aware of:
In short, Rainbolt is one of a slew of content creators who have essentially turned Google Street View into an vigorous esport. They play GeoGuessr, which is a game that plonks you into a random spot on the planet and invites you to drop a pin where you think you are. I remember playing GeoGuessr briefly about a decade ago, but in the years since they’ve included enough new features, multiplayer functions and game modes to make it a relatively robust platform.
If you check out Rainbolt’s TikTok account, you’ll see a startling array of videos which of him identifying his location on the planet very quickly and with unsettling accuracy. As the above thread points out, he uses all kinds of indicators to place himself on the globe, from the obvious to the esoteric. Architecture and the style of street signs are both standard, but he’s also very good at identifying the colour of soil and drawing inferences from Street View-specific information — like the implied height of the car’s camera.
I find people like this so fascinating for a couple of reasons. Firstly, it seems like just about any kind of competitive game environment inevitably produces characters like this, who are willing to apply their particularly odd brains to completely mastering it, even if the skill itself isn’t really that relevant outside the strict confines of the gameworld. (Unless you are kidnapped by some kind of global trafficking cartel and released into a random non-urban environment without a phone, being uncannily good at locating yourself by the colour of nearby soil is probably not super useful.)
But also there’s a broader idea of the metagame. A term emerging from game theory but now used quite a lot when talking about video games, it refers to the notion of thinking about a game outside its prescribed borders — essentially, using external information and reasoning beyond the game’s rules to win. This is why you’ll hear people describe high-level chess as a game within a game: yes, you need to understand how the game works on a deep level, but to run with the big dogs you need to understand how your specific opponent thinks, how they might respond to certain actions, and what the prevailing trends and strategies in the wider chess world are.
You’ll often hear online gamers talk about the meta when referencing big multiplayer games like Call of Duty, where they’ll essentially mean the bundle of strategies and assumptions that are most likely to win a match. It's the dominant way of playing, which is both generated by the way the game works, and the culture of the community that plays it. If the developer puts out an update which — for example — applies minor changes to the characteristics of certain weapons or maps within the game, it’s possible that could radically (and inadvertently) change the overall metagame. The world’s top-ranked players in any given title are always operating on the level of not only becoming very good at the mechanics of game itself, but comprehending the ever-shifting metagame and adapting their playstyle to exploit it.
People who excel at that very sophisticated strategic thinking but then apply it mostly to something as unremarkable as an online browser game are automatically fascinating to me. But GeoGuessr is unique in that to become good at the game you need to develop an absolutely insane knowledge of the tangible world itself. So, in order to become very good at the complex metagame of GeoGuessr, Rainbolt and others like him essentially become more informed about and attuned to our physical reality than basically anyone else on the planet. All to top the scoreboards on a website.
Isn’t that weird? I think that’s weird. (Bonus: here’s a video of Rainbolt playing GeoGuessr blindfolded, having someone just describe what they’re seeing on the screen to him.)
A good video
Web3 priority number one: try and get the normie podcast pitch down pat while the crypto economy is in the toilet.
An incredibly weird follow to the Google AI stuff from earlier in the week. The NYT reports that the Fellowship of Friends, an esoteric Californian sect that believes in achieving a higher consciousness through consumption of fine art and culture, has somehow infiltrated a business unit at the company, with a video producer complaining he was sacked for pointing it out. (This is obviously extremely my shit.)
On the woes of the buy now, pay later industry as interest rates start to climb and tech stocks fall.
From 2020, a story on a weird Cold War era radio station in the swampland outside St Petersburg that no one claims to run. “Twenty-four hours a day, seven days a week, for the last three-and-a-half decades, it’s been broadcasting a dull, monotonous tone. Every few seconds it’s joined by a second sound, like some ghostly ship sounding its foghorn. Then the drone continues.”
On ‘deep-fried memes’: “These images, crackling with yellow-white noise and blurred like the edges of a CGI ghost, evoke the distance between writer and reader on social platforms. Posts are refracted through filter after filter and pixels lost through screenshot after screenshot, singeing off the fingerprints.”
Enjoyed this on Blake Lemoine and LaMDA, which I wrote about earlier in the week. “Lemoine's problem is not that he has let his imagination run wild with his conversations. It's that he hasn't.”
I thought this was interesting from a cryptographer summarising recent developments in blockchain technology, and where he sees it going once the more hyped applications fade away or fail.
An interview with Pankaj Mishra on Russia and Ukraine.