This is a revision and expansion of two (bad) pieces I wrote; Platforms and Monetization and Slot Machine Galaxy. While I haven't progressed too far in studying this stuff, I am glad to say that I have progressed enough that some of my past writing is starting to embarrass me. The underlying ideas are still relevant, and I think I have some good material, but I want something a little more serious and definitive than those old treatments.
I originally wrote this in June-July 2024. I thought it needed a conclusion but I think it has some value as is. It is not completely thorough, and drops off in detail near the end, but it is an OK sketch of what I've been thinking about. You'll have to wait for videos if you want me to push these concepts into their full cosmic horror significance.
In a previous entry I looked at how economists treat digital commodities, and found that digital files are widely (and correctly, in my opinion) classed as non-rival goods. i.e.:
Copying is not theft,https://www.youtube.com/watch?v=IeTybKL1pM4
stealing a thing leaves one less left
copying, it makes one thing more,
that's what copying's for.
Files are non-rival because everybody can have a copy, and with widespread internet access there is no cost of reproduction. In theory this state of affairs hurts creators, but in practice the paradigm hurts publishers. Steve Albini (RIP), a consistent advocate for musicians and against publishers, described the vampirism of major record labels in this excellent talk. Suffice to say, the major labels aimed to extract as much profit from musicians as they could, often to the extent that bands never saw any money. This intense exploitation was possible because labels had a monopoly on the (physical) production, promotion, and distribution of music. Bands needed to get their music on the radio and in stores, and record labels possessed all of the connections. Independent labels emerged as an alternative, and were often a much better deal for bands. Then the internet happened.
Computers and the internet destabilized this old system because artists could not only produce their own records and distribute them online, but with the help of social media they could hyper-efficiently find their niche. The internet creates new problems for artists [1] but on the whole this system is much better for art as a whole--there are few barriers to remixing--and it is much better for less popular artists who want to make money. The Patreon model has proven that people are eager to support creators (or just people) who they like, and this epoch of non-rival goods shifts attention from product to creator. Again, there are problems with this, like parasocial attachments, but for many it means a creative career that can survive the inherent risks of producing an ambitious work of art.
Donations, later bolstered with ads, have more than paid for all the books I need to make Pay to Win. We're not living in utopia here, the time I've put in would be valued at a few cents per hour, but the economies of the internet are about as good as you can get under current conditions.
Old institutions wanted to assert their control over this new ground. The Digital Millennium Copyright Act was an important step toward doing so, but it was a defensive move to protect properties that the publishers already owned. The DMCA adapts the idea of intellectual property for the internet. It protects copyrighted works, but more importantly the DMCA makes it illegal to circumvent access control schemes "whether or not there is actual infringement of copyright itself". It secures a monopoly on access for anyone who can afford a lawyer, which major record labels certainly can [2].
At this point games split from other forms of media. For movies and music, the positive assertion of control came in the form of services like Netflix and Spotify. Not only was media out of the consumer's hands, but the publishers also reassert their control of distribution. You can't just put your music on iTunes or Spotify, you have to pay a service like DistroKid to do it for you. Spotify revenue is pitiful, even for fairly popular musicians, so most artists on Spotify are paying a tax for the possibility of being discovered. Major labels command most spots on Spotify's playlists, so there are also new incentives for artists to sign.
Access control is an old strategy (private property), but active DRM measures made it obvious. The NES lockout chip is an early example: Nintendo owned their hardware design, so your game would be unplayable if you refused to pay Nintendo's fees. If you squint, you might see a blurry picture of Steam's model here. The NES was a physical hardware-firmware platform, but the internet made the concept of a platform more ethereal: Roblox is a platform, Steam is a platform, Spotify is a platform. While these platforms may have different business models, they do exhibit some remarkably similar tendencies.
Steam resembles a traditional store, so it makes a good starting point. Steam launched to serve patches and host servers for Counter-Strike and Day of Defeat. When Half-Life 2 came out, Steam and an internet connection were required to install it (in 2004!), and apparently the launch was a bit of a disaster [3]. This was Valve's first big move in digital monetization, and it would not be their last. Following HL2, Steam started making deals with third parties, and became a bona fide marketplace. Today, Steam will list pretty much anything if you pay them a $100 fee.
Initially, I assumed there were major problems preventing games from fitting into the Netflix/Spotify streaming model, but this is not really the case. I had assumed that Google Stadia failed due to quality issues, but it reviewed quite well and simply lacked users. The service ran games on Google's machines, streamed the audio and video to the user, and streamed user input back to the machine. PlayStation Now (later rolled into PS+) launched in 2014 and does the same thing. Anecdotally I have heard that PS Now is both perfect and unusably bad. The model is untenable for people with bad internet, obviously.
The GamePass concept is more recent (2017), and it is not clear if it's profitable (Phil Spencer has said that it is). GamePass offers a catalogue of games for a monthly fee. You still have to install them on your machine, but GamePass acts as DRM. This is a good deal for the consumer but it inherits all of the downsides of Spotify/Netflix plus the issue that game downloads can climb into the hundreds of gigabytes. Microsoft owns the service, and I wonder if their aggressive acquisition strategy in recent years is part of the plan for making GamePass successful.
Games are expensive to produce, and major publishers sell them at $70 per copy, so they have little incentive to sign exclusivity deals with some particular service where they will lose out on sales and be subject to an untested revenue model. Microsoft won't even make its own properties GamePass exclusives, because they're not stupid. If AAA fatigue really sets in, the GamePass model could potentially become popular as people stop buying games at full price, but the industry is still limping along at time of writing [4].
This is all to say that Steam's storefront model proliferates right now. Big publishers are starting to experiment with streaming-like models but none have taken off yet. I am inclined to say that a switch to streaming is inevitable as the media industries shift to retention-focused strategies [5], but it is possible that existing marketplaces like the App Store and Steam will distribute free-to-play live service games and collect rent from the microtransactions.
Steam's model is quite simple. Summarizing from their own documentation: Developers pay $100(+tax) to submit their game via Steam Direct. If it is approved Valve takes 30% of sales, while the developer (or publisher, depending on the case) gets the other 70%. The $100 fee acts as a deposit, recouped after the game accrues $1,000 in gross revenue. The use of "recoup" in the documentation is very specific but it does not seem to be a trick.
The $100 fee practically guarantees that Steam will profit from every submitted game, because Steam operates at an immense scale. This is not difficult to demonstrate.
First I have to acknowledge that Valve is a private company so we do not have access to any numbers. But we can establish an upper boundary and work from there.
First, we will calculate the cost of labour. levels.fyi reports a median salary of $202,571 CAD (=$148,836.39 USD) at Valve, Zippia reports $68,007 USD. Glassdoor skews toward the former number though the $150,000 figure is still quite high. A 2016 source says the company had 360 employees, so let's bump that up to 500. To cover my bases I will bump the labour cost to a generous $200,000/employee, before equipment.
L ~= 500 x $200,000 = $100,000,000/yr
We'll add another $20,000,000 per year for equipment and office costs. This is more than reasonable; at $50 USD/sqft/yr, Valve's 135,500sqft office should cost around $6,775,000/yr. That leaves thirteen million for overpriced chairs, computers, and any on-site facilities. This cost will be F.
L + F = $120,000,000/yr
Due to all of my assumptions this figure should more than cover the equipment and labour required to run Steam, the estimates run very high.
Steam's largest cost is bound to be file serving. They need a ridiculous throughput of around 20 terabits per second, with double that at peak times. They use mostly Akamai, some self-hosting, and some other companies, but specific numbers are not available. In lieu of those I am using the consumer pricing for Microsoft Azure's content delivery service. Valve is almost certainly getting better prices than this, since cost tends to decrease with volume, down to some absolute limit.
20 terabits is 2500 gigabytes, and there are 31,536,000 seconds in a year, so Steam handles approximately 78.84 billion GB per year. Azure costs $0.028 per GB, so hosting costs should amount to $2,207,520,000. This is an intimidating number, and I suspect the real cost is quite a bit smaller. Amazon Cloudfront is still $0.02/GB if you are handling 5+ petabytes of data per month, but they offer discounts for guaranteed traffic, and Steam can guarantee that 20 terabit per second base load. Either way, our hosting cost will be $2.2B per year.
This gives us some insight into why Netflix has to scramble for profitability while Steam is a stable source of income. Steam's model extends the older pay-per-game system, so Valve has already recouped the cost of serving a game before the user can download it. In a direct streaming setup different users will have different demands, and if a given user's watch time goes beyond some threshold, Netflix will lose money on that user. They have to serve more data than that subscription pays for. Steam self-regulates its hosting costs; you cannot make use of the content delivery network (CDN) until you've paid for it.
They do have to swallow a minuscule portion of the cost to serve the actual Steam store and handle servers for games, but the lion's share of Steam's CDN is tied up in serving AAA games full of uncompressed sound effects. I will disregard the web hosting costs.
So, why does the $100 submission fee guarantee that Steam profits from every game? The cost of storage is extremely small compared to the cost of hosting, so even if my game never sells a copy, Steam makes money.
If I do sell copies? Let's assume that I make a game and sell it for $8. Valve takes 30%, or $2.40 per sale. At our quoted price of $0.028/GB, my game would need to be over 85GB for Valve to lose money serving it. Every GB over 85 will eat $0.028 of the submission fee per purchase, but this is an absurd situation. High quality (i.e. high volume of downloads) indie games rarely break 10GB, and $8 is an extremely low price for such a game. $25-$30 is the norm. An 85+GB game at such a low price is bound to get rejected by Steam Direct, or is some totally unoptimized asset flip which will not sell in high enough numbers to cause a problem. Even if some games manage to lose money--free games that become hits, for example--the costs are recouped through other high-profit games on Steam. Steam's size makes it more stable.
Take Dwarf Fortress, for example. It costs around $30 and weighs in at only 500MB. We know that it has sold over 800,000 copies, so 800,000 will be our sales number. Valve makes $9 per sale, minus $0.014 to serve the 0.5GB download. That's a profit of $7,188,800. The cost of serving the 400 total GB of dwarf is $9200. Across Steam's entire library of games, this ridiculous profit margin creates a reserve fund that can eat any costs created by Steam Direct's open door policy. Basically, the rich get richer: hit games can subsidize the process of finding new hits.
You can imagine how ridiculous these numbers get with a AAA release, even if Steam gives major publishers a larger cut. Steam's famous sales reduce the per-unit profit margin but drive a higher volume of sales. Again, any disappointing sales are subsidized by successful ones.
Even if we assume that the peak player count is the entire install base, most games are profitable. Diablo 4 has a peak player count of 39,599, is a 90GB install, and costs $60. Serving the game costs Valve around $100,000, but they make $700,000 (gross) in return. $15.48 net profit per install. In truth Diablo 4 has probably sold hundreds of thousands of copies on Steam.
Beyond numbers, the point I am making here is that Steam's model incurs almost no risk. If Diablo 4 sold one copy, Steam would profit from it. If you've tried to play D4 on Steam you might be reminded of the game's frequent, ridiculously large updates. A user would have to download Diablo 4 eight times for Steam to lose money. Even if that happens rarely, most users will not do this so the cost is covered by Steam as a whole. Given Steam's de facto monopoly on game distribution they are unlikely to fall into the red any time soon.
Although Steam is analogous to retail sales, Valve got in early to establish its dominance, and Steam has been able to adapt and become the site it is today. Through their dominance, I'm sure Valve had a hand in making digital games cost the same as physical copies, even though digital games don't require physical production or distribution.
It is easy enough to understand Steam today, but the service has a set of contradictory incentives, and it needed to overcome these to become the true, stable juggernaut of game distribution. It is in their interest to run the site with permissive policies, to 'automatically' capture niches and guarantee that they benefit from unicorns like Among Us, Lethal Company, or Helldivers II. Listing as many games as possible caters to as many consumers as possible. Steam's review process basically just checks that submitted games are not viruses or illegal, there's no quality review in an artistic sense. Steam does not care if your money is going toward a good or bad game, only that 30% of that money goes to them. The site's official position as a store for every type of gamer [6] says the same thing implicitly.
Over time this policy creates a major problem: there are more bad games on Steam than you could play in a lifetime. More asset flips, achievement generators, demand generators, and first attempt RPG Maker projects than a healthy marketplace can absorb. The market's arteries become clogged, and it takes longer for consumers to find games they want, so some people give up, and sales are lost [7]. The store also "looks bad" overall, so people are hesitant to use it. Sites like G2A that sell cheap Steam keys are a good example, they have an ick factor that needs to be compensated for with lower prices. It goes without saying that the profit motive creates these bad games, asset flips only exist because they sometimes make money.
I have to stress that Valve is exceptionally competent compared to other tech companies, and they are capable of taking a long view of things when the need arises [8]. Steam Greenlight, launched in August 2012, was Valve's first attempt to balance the quality and volume of games on the store. At launch, anyone could submit games to Greenlight, and users would vote for the games they liked. If a game was sufficiently popular (and, presumably, not a virus) Valve would list it on the store. Greenlight submission was open, so it was quickly overwhelmed by spam and Valve implemented a $100 fee (which initially went to charity).
Greenlight was still editorial in nature. Valve was selecting games that they thought would be successful. Consequently, the program was slow to adapt to changes in taste and Steam was still potentially missing out on hit games, since hits can't really be predicted. They released batches of games quite frequently--around 75 per month, looking through the old announcements. This was not a high enough velocity for Valve, though.
Steam Direct replaced Greenlight. It is the current system for indies to get their games on Steam and I described it earlier--$100 fee + 30% of revenue. The fee and review process have maintained a minimal standard of quality on the store but Steam Direct removes the editorial aspect of Greenlight, so it can capture every game that could possibly succeed. Steam Direct creates a high volume of new games, but it leaves the site without a way to privilege good ones. All else being equal, good content becomes impossible to discover if the mass of bad content grows too large. Just look at Twitter.
This discoverability problem arises on all large digital platforms (Netflix, Spotify, Twitter, YouTube, Amazon...), and the solution tends to be a recommendation algorithm. Steam provides a smattering of different solutions: curators, reviews, the discovery queue, and a simple recommendation system. The recommendation algorithm is closed source, but inferring from the sidebar on all store pages it performs simple checks: does this game share a lot of tags with games in the user's library? Do the user's friends and followed curators like it? Are the reviews generally positive? Some weighted average of these determines the probability that the user will receive a recommendation.
In my experience, Steam's algorithm is quite good, but because I follow many game developers on Twitter it rarely serves up anything new. Occasionally it will show me a remaster or rerelease that I did not know about. They also have an interactive recommender which apparently uses machine learning. Quality-wise it is indistinguishable from all of the other systems.
The fatal flaw of all recommendation systems is that they can only offer more of the same. If we take Google seriously, a large enough mass of data may model a more abstract thing like personality or taste and actually recommend works that are both new and relevant, but I will talk about that if and when it happens. This 'more of the same' issue is very prevalent on Steam, because there are way more games than there are interesting ideas, and games are expensive. A quick check of my discovery queue served up some popular mainstream games, some survival-craft games that I'm sure are worse than Valheim and some automation games that I'm sure are worse than Factorio.
All recommendation systems have inherent qualities, they favour one data point or another, and this is true whether the system is a string search (favouring the string closest to the input string) or a hyper-complicated machine learning system [9]. In fact, black box systems have huge potential problems because you can't mitigate a disaster until it happens. You can check if a text search works, and you can check if the user can exploit it with SQL injection or some other method. These fancier systems are called black boxes because they cannot be checked.
The scandal of YouTube kids videos is probably the biggest disaster involving a recommendation algorithm, or at least the most memorable. Because YouTube uses machine learning to recommend videos, no one could have predicted what exactly the algorithm would fixate on and distribute to kids, but it ended up being a bunch of surreal mixtures of super-heroes and Disney princesses getting pregnant, getting injections, shitting themselves, miscarrying, and killing each other. Even if the content was not reprehensible on its face, it has the character of slop. These videos were only created to game the algorithm, and they are only watchable if you have the extremely malleable brain of a young child. They lack a discernible plot or characters (in a literary sense), and instead comprise a soup of shocking events featuring characters (in a "hey look it's Spider-Man" sense) that children might recognize.
There is a subtle horror and discomfort seeing these symbols reheated in somebody's garage or a modded copy of GTA4, even as an adult. It's a little bit Lynchian, the image has some familiar stuff but it's arranged and exaggerated into a dream. There's none of David Lynch's overt horror, though. Algo-slop is scary because it is a machine's dream, imagining the audience and purpose just reveals an abyss. It's not hard to explain why Else-Spiderman happened but the specifics of the content have a very indefinite relationship to human nature. Would a 'perfect' algorithm turn into this too? Is it us or the machine?
If a platform is a system which takes in content and spits out recommendations, the algorithm (transfer function) prescribes the incentives for creators. YouTube's algorithm is so complicated that it can only be gamed in a very general or transient way. Elsa-Spiderman videos don't specifically work anymore, but in general maximizing watch time is the key to success through YouTube's monetization.
Returning to Steam, in the last few years many independent developers have started pushing for players to wishlist their games before release. This has all of the obvious effects [10], but I have also heard from developers that wishlisting boosts the recommendations at launch. Steam's algorithm changes, and this changes creator behaviour in turn. This reading of the algorithm by its subjects does not need to be true, but in this case wishlists do boost overall visibility [11, FAQ question 6].
I often come back to YouTube as an example because, by a series of accidents, YouTube creators are absolutely shameless in chasing algorithmic success [12]. There is a clear feedback loop at play between creators and the algorithm: somebody finds a new idea that becomes a hit, then other creators pick up on it and reproduce it until its popularity saturates and they move on to the next thing.
Regular (i.e. pre-internet) trends work in similar ways, but algorithmically-driven sites have a much higher volume and velocity than trends in, say, mainstream film-making. Internet memes are truly endemic to the internet, since digital files are uniquely easy to remix and the internet offers frictionless distribution. If we want to take up the memes/genes line [13], genes only change over the course of millions of years, means of communication on the internet evolve in minutes.
If we abstract from reality a bit, this speed is unique to the internet but not a priori bad. In a world with stronger social bonds, where we managed to maintain a barrier between online and IRL, internet memes would be harmless. Things did not work out that way, but I wouldn't rule out a mass exodus from the internet either.
Because people are using these high-speed algorithm-driven platforms to make money, the market quickly fixes on effective strategies to get clicks and views. Content becomes grotesque in its exaggerations but goes down smooth, lacking any surprise or challenge. See the ultra-rich sex pest Mr. Beast. Genuinely great work can still succeed but only if it is sufficiently close to something that has succeeded before. Vivziepop's work, for example, is not quite Dostoyevsky but it is genuine. At the same time the speed, bright colours, and flashy designs in her projects are probably a major factor in her success. They are both bright and edgy in a way which is perfect for viral success, even though the product is probably genuine. With that said, truly challenging work is still better off than before the internet [14], because many more people can find 1000 true fans [15].
On YouTube the border between genuine and cynical work is fuzzy. A successful strategy can be reproduced and copied so quickly that the simulacrum is practically simultaneous with the original {++ reading this back in 2025, that's kind of the whole point of the simulacrum concept--there is no seam where the simulacrum connects to the real. ++}. This is sort of a formal distinction, because love has no intrinsic relationship to video quality, content, or discoverability, but a video is a very complicated set of things. Someone who doesn't care enough to consider their decisions in writing, performance, or editing is likely to produce a bad video. If that video gets ten million views, the problems are replicated and exaggerated by the next person, and so on.
{== I started writing this months before these videos became a meme btw. The gas leak in my house has me feeling like an oracle. ==}
Here in Parkour Civilization, no one chooses to jump for the beef.
This is Mining Civilization. Mining determines where I live, how much food I get, and how people treat me.
I live in the Dirt Civilization and every single day the dirt officer comes and collects dirt taxes.
This is Redstone Civilization, a place where players are trapped in the confines of this wall.
I only have two swings left of my sword until my life in PVP Civilization is over. But that's how PVP Civilization works.
This rapid fixation on successful strategies happens on many axes at once. Certain sound effects are reproduced in all of the videos I quoted (vine boom), and Evbo's success popularized a particular style of thumbnail which appears in videos beyond the X Civilization genre. Even the names of these YouTubers are fairly homogeneous, something I first noticed when people started calling themselves "thomas is not at home" or--and I just made this one up--"dream was taken."
Successful new strategies are copied and degraded so quickly that the original is lost in its own wake [16]. Viewers who were 'there' for the original may notice this happening, but a greater mass of viewers will not, they will just find this new genre in situ. YouTube's measure of creator success is previous success, so quality original content is not privileged over derivative content. In point of fact, original content is riskier and often rewards creators less, financially, than chasing trends.
Google search was ruined, in part, by the effectiveness of SEO spam, which could be produced in much greater volume than genuine content. AI is seen as an existential threat to creators because it allows people to apply SEO spam logic to audiovisual works, outcompeting effortful work by drowning it with noise. For most people the only way to interface with art is through an algorithmically-driven platform, and making a living from creativity requires you to meet your audience where they are.
I am comfortable stating a rule from this observation: the discoverability problem tends to become more intense on platforms with a lower barrier to entry. Games are quite difficult to make, videos are quite easy to make. Twitter is famous for being awful. The profound "democratization" of art with AI hurts anyone who puts passion into their work. It even hurts those who make interesting things with machine learning [17]. This can be called a problem with capitalism in the sense that the lion's share of cynical art is motivated by profit [18]. On the other hand, I will also remind you that Silent Hill exists because Konami wanted a bite of the Resident Evil market, so things really aren't that simple.
A new consumer appears at some point in time, and they do not see the birth of trends but enter a platform where trends already exist. It is impossible to discern quality among the quantity, so the viewer either dismisses the genre for its abundance of low-quality videos, or accepts the generally low bar and follows whatever creators they are drawn to.
All creators are subject to two deeply related factors: algorithmic success and content quality. Some quality is required to retain an efficient number of viewers [19], and specific qualities of the video will determine its algorithmic success to an extent. A creator with previous success can drop production quality and still retain viewers. If the aggregate level of quality is low, viewers have no way to discern quality (without investing their time), and viewers are likely drawn to creators by things like their voice or personality in the first place. Quality is not a major factor at the top.
I conceive of 'quality' as a combination of technical prowess (in terms of recording, editing, writing/sociability of the creator...) and originality. This is not a good definition, but it is sufficient. We're on the precipice of ethical philosophy here so I don't want to go deeper.
From a cynical creator's view, the discoverability problem is an opportunity for low-effort success. They can invest effort into tracking public opinion and hot topics within the algorithm, and copy successful content cheaply. Since quality cannot be discerned ahead of time, the name and thumbnail will inevitably generate some views. All else being equal, a low-effort creator will outproduce a high-effort creator, giving the low-effort creator more opportunities to score a hit within the algorithm. Once one video hits, the next is more likely, and so on. SunnyV2 logic.
Slop is an appropriate name for this sort of content not only because it is a mishmash of algorithmically successful strategies, but because it creates a lot of "dead" views. Viewers who clicked on the video expecting something effortful leave unsatisfied (i.e. they go watch another video) but still view the ads and bolster the creator. These videos are the slop/slack/play in YouTube's system.
The platform functions as a slot machine from the creator and consumer perspective. Hits reinforce the behaviour (scrolling or producing), and everyone involved becomes more tolerant of losses. General quality declines. All markets are slot machines.
The platform is incentivized to promote low quality content until the fatigue of watching slop causes users to leave. A laissez-faire approach like this is unlikely to create a culture, all creators are competing in a race to the bottom [20] and users are frequently tricked by slop. There is no room for trust.
Gatekeeping is not nice but we must acknowledge that a community is defined by who is not in it. Heavily editorialized sites may not be fair but they do offer a common set of ideas that users can chew on, discuss, and develop. For someone reading this the word Pitchfork probably summons up a very specific time, place, and list of bands. When the great media consolidation was happening, it seemed like a natural choice to shift from self-hosted video to YouTube, or from a blog to Twitter, or to let Conde Nast buy your website. You can reach a bigger audience this way, after all. It turns out that the incentives of publishing dissolve these communities and voices. As I've said several times, having character will always alienate someone, but what a strong voice gives up in market value is more than made up for in its social value. Unless it's Elon Musk's voice or something.
A platform is a marketplace owned by a third party. Exchanges are not necessarily made with money, but platforms that facilitate traditional purchases (e.g. Steam) entail very little risk for the owner. Successful platforms tend toward monopoly. Platforms are incentivized to maximize product volume. Consequently, low-quality products flood the market and some mechanism is needed to facilitate sales, in other words to connect people with the products they might actually buy.
The ethereal new concept of a platform is clearest in a game like Fortnite, and it pioneered a successful monetization model [21] that does not fit neatly into the Steam-Netflix binary I have worked with so far. Social media sites usually monetize with ads rather than subscriptions but they use the streaming service format. Low or no-cost access is needed because the users are the product and the site has to maximize volume.
When it became popular, Fortnite was a live service third-person shooter with a shop. It was an unimaginative spin on the 100-player Battle Royale, and a huge hit with kids. It was released as "early access" in late 2017. In early 2019, Marshmello "performed" a live set, and from there the game shed most of its character and transformed into a platform. It seems like ancient history now, but Fortnite was a 'metaverse' contender. The game's synchronous events, like the Marshmello set, echo the live service model; Fortnite is split into periodic seasons and chapters that offer new cosmetic and map content along with a new battle pass.
Today Fornite is a battle royale game, a survival game, a sandbox game, and a rhythm game. All separately monetized, of course. Early in Fortnite's life Epic offered an affiliate marketing program (their words [22]); content creators got codes which their fans could put into the game. When one of those fans made a purchase, the creator would get some money too. This style of marketing is good for acquiring customers, but to become a platform Fortnite needed to drive the creation of high-quality content.
They do this by paying users who create successful games (called "islands"):
"Engagement payouts proportionally distribute a percentage of the net revenue from Fortnite's Item Shop and most real-money Fortnite purchases to all island publishers on a monthly basis. Payouts are based on what players enjoy in Fortnite and account for island popularity, engagement, attracting new players, and other measurements..."
From [22].
Fortnite is probably the most familiar example to you, but Roblox started a similar program all the way back in 2014 [23]. It did, indeed, drive the creation of high quality content. Roblox was, from its inception, what Fortnite is now trying to imitate: a platform where the players produce an endless stream of content and revenue. Roblox has always been openly financialized, too. Back when the game had two currencies there was a currency exchange (RoblEX [24]) with rates driven by the playerbase. Players could buy ads on the site, and limited-edition cosmetics are traded exactly like CS:GO items.
With the addition of the Developer Exchange in 2014, Roblox gave RBX an exchange rate in USD. Today, games can have special passes priced in RBX, and subscriptions priced in dollars. These don't apply to the whole site--just one game on a site that publishes hundreds of thousands. Players can pay to make private servers. Players can publish the most varied kinds of cosmetics, which they can sell for money. On one side, the customers are children with no impulse control. On the other, the developers are children perfectly willing to overexploit their audiences.
Roblox collects rent on each transaction, and the game's high-quality user-generated content ensures a long lifespan with a high volume of transactions. The company is protected from responsibility, since the scummy practices of individual creators appear as isolated incidents, even as Roblox reaps the rewards.
Like Fortnite and Steam, Roblox has shed its anachronisms over the years. Characters now look blandly human by default, and most accessories are user created, whereas the shop used to reflect the voice of the development team (as strange and Silicon Valley as they were). Even the games, which were once called Places, are now ethereal Experiences.
Fortnite is copying the culture and look of Roblox; they have a Lego partnership and their Twitter page promotes "tycoon" games [25] and obstacle courses, two of Roblox's oldest and most characteristic genres. According to fortnite.gg, most Fortnite players are still in the official games made by Epic. While these may be better works of game design than the average Roblox game, Fortnite Battle Royale has an expiry date while Roblox generates new content forever. Every time I look at Roblox, the most popular game is different.
The metaverse extends the internal logic of platforms. If a platform is set up correctly, it generates rent essentially for free. Steam's product is its monopoly, they are selling developers the opportunity to be seen and the legitimacy of a spot on Steam's storefront. There is no point arguing about the value proposition because Steam has no competition. They can take 30% of sales if they want to, so long as the perceived cost of leaving Steam is higher than the cost of staying.
Platforms seek a monopoly, and companies struggling toward the metaverse were trying to enclose all digital life in their fee structure. Second Life is instructive, but I will stay with Roblox.
As I mentioned above, Roblox has a 'platform layer' of monetization through RBX, which players buy for real money. Roblox takes all of this money and binds it into the game's economy. Within the platform, there is 'content layer' monetization in the form of accessories and game-specific passes or subscriptions. When these are priced in USD I imagine Roblox takes another cut. Simply because Roblox owns the platform, they can take a huge cut from any user that tries to exit the RBX economy as well, about 60% [26].
In the minds of Californian CEOs, ownership of the metaverse would allow them to collect fees on practically every purchase the user makes. Every hypermonetized game the user plays, every grocery purchase, every Uber, every cs_office.bsp a company rents for a meeting...
At the level of NFT speculators, the metaverse craze made some sense. The harder pill to swallow is that Facebook tried to move on this blatantly stupid idea. The only explanation I can come up with is that they had dollar signs in their eyes. Even if VR were affordable or comfortable there are only so many asset flip Unity tutorial virtual experiences a person can put up with. VRChat is the closest thing to "the metaverse" and its approach has been radically hands-off (and, importantly, pornographic).
The real, mundane version of the metaverse is probably some form of augmented reality. If AR glasses become cheap and ubiquitous, an employee with AR could outperform one without (in a fake office job). The real promise isn't a totalizing new fee structure for Meta, but an endless canvas of ad space tied to hardware the user cannot easily mess with. Facebook and Google are ad markets first, and their products are subservient to ads. According to a Wired article from 2020: "Google earns more than 80 percent of its revenue from advertising; Facebook, around 99 percent" [27].
There is no agreement on whether ads actually work. Tim Hwang, the subject of [27] wrote a book arguing that digital ads are not effective and most clicks are fake ("fraudulent," in his words). Subprime Attention Crisis draws an extended metaphor with the subprime mortgage crisis, implying that the tech-advertising nexus is a bubble. The likelihood that advertising works at all is questionable in an era as cynical as ours--I have only ever seen ads work on elderly people and small children, so in all likelihood advertising's main effect is to hurt weak people. Articles and books abound on, for example, ad executives' narrow definition of beauty, so any consumer with time to think about this can discount ads as a source of reliable information.
But, like that strain of feminism that seeks to make women fully working class, rather than fully free, any push to redefine beauty has happened within the advertising superstructure. Broadening the mainstream concept of beauty comes from a nominally good place, but ads will still try to make the viewer feel ugly, no matter what content fills the Perfectly Beautiful Person mold. Liberal egalitarianism is about making ads where everyone is free to feel ugly and dissatisfied--in fact, it generates a deeper sense of ugliness beneath the skin.
Anyway, acknowledging an ad bubble raises the concern that this bubble will pop. It could happen, but I have serious doubts. Even if ads do not work in specific, they present a normative vision of the world and are useful for shaping people in aggregate. I don't think there's a conspiracy, but ads promote consumption in general and I think there's an acknowledgement that ads have some kind of unquantifiable value.
[28, par. 2]"Reality considered partially unfolds, in its own general unity, as a pseudo-world apart, an object of mere contemplation. The specialization of images of the world is completed in the world of the autonomous image, where the liar has lied to himself. The spectacle in general, as the concrete inversion of life, is the autonomous movement of the non-living."[28, par. 6]
"The spectacle grasped in its totality is both the result and the project of the existing mode of production. It is not a supplement to the real world, an additional decoration. It is the heart of the unrealism of the real society. In all its specific forms, as information or propaganda, as advertisement or direct entertainment consumption, the spectacle is the present model of socially dominant life. It is the omnipresent affirmation of the choice already made in production and its corollary consumption. The spectacle's form and content are identically the total justification of the existing system's conditions and goals. The spectacle is also the permanent presence of this justification, since it occupies the main part of the time lived outside of modern production."
The Spectacle plays the role of Starscourge Radahn--it yearns for the end of history, and represents (but merely re-presents) total closure to revolution (or meaningful reform). It wants to hold back the stars, and it wants to make this holding-back reflexive. If an ad does not present you specific new needs it at least reinforces the concept of scarcity. I also get the feeling that advertising works in specific. We know Instagram is very good at giving teenagers eating disorders [29], sponsored content probably works too. Facebook is also slowly dissolving the brains of the elderly. Again, this only covers children and seniors, but there are gullible people out there.
This is to say that advertising is probably too useful to simply die out because it doesn't work in an easily quantifiable way. Ads condition the proper reflexes for an economy driven by consumption. Advertisers would flee Google and Meta if they began to hemorrhage users, but not before that point. It is obviously important to these companies that you see ads, even if the cost per impression is questionable. Even the process of showing ads to users runs as an autonomous marketplace; Google's ad technology is a black box but every time a page loads a Google Ad, advertisers are participating in a silent, instantaneous auction for the space [30]. Forcing advertisers to compete maximizes the CPI, and Google's take.
Creators who benefit from advertising have the least power in this regime, since the market-maker and advertiser both have more resources than a creator. Losing monetization is a big deal, so the average creator must be intensely conservative; all swearing is censored, anything questionable is communicated through euphemisms, and content can never discuss contentious social issues. Billion dollar companies love "the gays" but only as a spectacle; completely sexless, bloodless, and distorted into flat design and pastel colours. People without humanity, literally reduced to tokens. Those are my people, but you too have been boiled into a soup of audience tags.
At this stage of privatization, platforms have replaced social institutions; public opinion is shaped and filtered by for-profit companies. Audiences love controversy, but the platforms select only for controversies that do not impact the stability of the platform or the greater ecosystem that allows it to exist. It is easier to feel the squeeze when I talk about social networks incentivizing you to say "unalive", but hopefully it is clear that these platforms are all marketplaces. Independent artists are now subject to the same glamorous market forces you might see at the grocery store. Will the new chocolate-flavoured yogurt succeed or will cappuccino win out?
These platforms are structurally reactionary. They have found success with a formula, and the creators they collect rent from succeed within that formula. Any deviation is automatically met with resistance. On YouTube, flashy thumbnails and clickbait titles perform well. Being honest automatically reduces your exposure. Of course, there are more impactful controls too, such as the aforementioned use of "unalive" or the fact that there can be no frank discussion of topics like sexual assault on most sites. The ongoing genocide of Palestinians would also be classed as controversial, and if you support Palestinians in a way that would actually change things--that is, if you support Hamas--you are openly breaking the rules of most social media platforms. By taking away income or reach, the platform automatically marginalizes victims, and scribes a line for what is or isn't OK to talk about.
The best and/or most successful creators have a lot riding on whether their videos are safe for ads, so they are heavily disincentivized from touching difficult topics. Platforms have a standard of discourse whose specifics are hidden, and anyone who deviates from this standard is threatened with demonetization or banning.
Forums of yore wanted civil discussion, so they published their rules openly. Platforms do not do this. The presence of ambiguous rules conditions a conservative reflex. Since YouTube won't tell you its stance on the genocide, the risk of talking about it is very ambiguous, and it is in creators' financial interest to over-correct and be as safe as possible. This extends to the mundane as well, just look at those Minecraft civilization videos I quoted earlier. People stick to what works.
The only benefit of tackling a controversial topic is spiritual, and few people have the privilege of valuing what's right over what pays the bills. Anything meaningfully new--whether it be an artistic expression or political movement--will be controversial. Platforms are inherently reactionary, in that they punish anyone who destabilizes their hermetically sealed worlds. Platforms seek monopoly, so they rarely acknowledge any territory beyond their own unless it is ripe for colonization.
In the case of social media platforms, this is a function of the profit motive. Unless this new movement has so much momentum or public support that it cannot be ignored, the platform will ignore it. The mainstream reaction to the genocide is actually instructive here. At first, the media strategy was to ignore the genocide, then to paint supporters of Palestine as bigots. Now that Israel's atrocities are more-or-less undeniable, institutions are trying to assimilate and destroy the destabilizing elements of the pro-Palestine movement. By repeating the word "ceasefire" until it is meaningless, or urging people to listen and have compassion rather than try to stop the genocide their taxes are paying for.