5 Video Game Things I Learned | 2026 August 8 - August 14
Boy, I wrote up a long one with lots of ranting today. That’s why it’s out so late. Lots of failures in these stories. More GTA VI leaks lead to lots of internal disappointment at Rockstar. Meanwhile, PlayStation continues to find itself banging its head against the live-service brick wall. I’ve got some AI-related rants in here that probably go on for too long. I then cap everything off with a mild story looking at how games get into big events like The Game Awards and Gamescom’s Opening Night Live.
PlayStation Keeps Failing at Live-Service
PlayStation can’t seem to get out of its own way with its monster push of live-service games over the past five years or so. It’s pushed its first-party studios that were built on single-player games to pivot to live-service. And the results speak for themselves. Well over a dozen failed games (most of which never saw the light of day), including Concord and games in the God of War, Twisted Metal, and The Last of Us franchises. They also have a whole debacle going on at Bungie where one of those live-service games was canceled, Destiny 2 development ended after an up-and-down nine-year run, and Marathon released to collective groans both from people not enjoying the game and those who felt squeamish at the controversy of the game containing plagiarized work.
This week, another game was added to this list of failures: Horizon Hunters Gathering. An upcoming co-op action game set in PlayStation’s Horizon series, Horizon Hunters Gathering was meant to be yet another attempt at a live-service game to unlock infinite money for Sony. And it is now yet another failed attempt at that. This week, it was revealed that due to poor player feedback, developer Guerrilla Games has spent the past couple of months ripping out all of the live-service content.
Through PlayStation’s live-service experimentation, the only one that has seen any sort of noticeable success outside of its already-existing MLB: The Show franchise is Helldivers II, a game where PlayStation only acted in a publishing capacity since it was developed by a third-party studio.
I think it’s pretty easy to see in retrospect why the PlayStation 5 has been such a quiet console generation thus far. The biggest successes for PlayStation since the PS5 released have been games like Marvel’s Spider-Man 2, Ghost of Yōtei, Astro Bot, Death Stranding 2: On the Beach, God of War Ragnarök, and Saros. One of those six was from a third-party studio, and four of them are from studios that, as far as we know, haven’t been bogged down by trying to create a live-service title for Sony. Meanwhile, other studios that have put out successful games for Sony in the past have found themselves completely stalled by live-service development that went nowhere. This includes:
Bend Studio (Days Gone) - was working on a now canceled live-service game and subsequently had 30% of its staff laid off.
Geurrilla Games (Horizon) - is actively working on the aforementioned Horizon Hunters Gathering that is at risk of being canceled if they can’t hit a milestone by December after having to remove all of the game’s live-service systems due to player feedback.
Naughty Dog (The Last of Us, Uncharted) - hasn’t released a new game this console generation despite it being almost five years old, largely due to them spending several years and hundreds of millions of dollars to develop a now canceled The Last of Us live-service game.
Bluepoint Games (Demon’s Souls Remake, Shadow of the Colossus Remake) - despite tons of experience remastering and remaking PlayStation titles to great success, was tasked with developing a God of War live-service game that eventually got canceled and led to the studio being shut down in 2026.
London Studio - PlayStation’s internal VR team that was releasing at least a new game every year, on average, was suddenly tasked with developing a multiplayer game playable outside of VR and was shut down in May 2024 after going about five years not releasing any titles.
And this list doesn’t even mention studios that hadn’t produced a game yet or released less-than-impressive games in their history, only to be waylaid or canned by PlayStation after working on a live-service. The most notable of these examples is of course Firewalk Studios with their game Concord.
It’s hard to imagine just how big of a hold PlayStation would have on the industry and just how successful its internal development teams would be right now if PlayStation didn’t put so many of its chips on the live-service model.
The Video Game Industry Is Shifting
Okay. This is pretty much always true. However, I think in the past 5-10 years, there’s been a dramatic shift that affects all facets of the industry, and that’s what Bain & Company refers to as “the average player.” Bain & Co. released findings from their 2026 Gaming Report this past week, and it focuses heavily on how the industry is so saturated that it’s sort of creating separate genre bubbles of the market to a degree that hasn’t been seen before.
Not long ago, the games industry was less digital and had far fewer games released every year. For reference, let’s look at Steam. In 2021, 11,217 Steam games were released. In 2025, just four years later, that number nearly doubled to 21,336. 2025’s numbers are also more than 7.5 times the number of games released a decade prior in 2015.
The other component of this that we can largely thank digital gaming for is the growth of the back catalogue. As more and more games release every year, the games that released in previous years don’t go away. I personally am very familiar with this. As of writing this post, I’ve managed to complete 24 games so far in 2026. Of those, only seven released this year. And this is a good year for me! In any given year, I usually only play three or four games that released that same year. There are so many good games from the past that I never got around to playing that are being sold at very reasonable prices.
This type of growth leads to an industry in which, according to Bain & Co., “The game shelf has become effectively infinite, and choices are multiplying faster than any player’s attention can grow.” So when Bain & Co. talk about the average gamer no longer existing, they’re talking about the new world we live in, in which game companies can no longer seek to serve large sections of the gaming population. While some games exist that can manage to capture that type of audience, targeting it is actually far too risky. Another common way of putting it is the age-old wisdom that if you try to make a game for everybody, you end up making it for nobody.
Since gamers don’t have to make do with whatever game genres happen to be serviced in a particular year, they can feel free to live within a specific niche of the industry that they really love, rarely venturing out to other genres or themes. This is why defining your target audience is so important. If you have a particular subset of the gaming market in mind, you can market to and service that market much better than trying to satisfy everybody. This gives developers freedom to experiment within their genre and avoid creating watered down features that don’t appeal to anyone.
The importance of catering to a particular audience becomes more apparent when looking at the results for play time and spending amongst gamers. The survey suggests that nearly 60% of total game hours played come from the those in the top 20% in terms of play time. Furthermore, 73% of video game spending comes from the top 20% of spenders. In other words, when you’re able to find your niche and cater heavily to the top 20% most engaged players within that niche, you’re bound for success.
What’s interesting is that this is largely how the world of books works. While I’m certainly not an expert in it, most authors have to spend the time finding and catering to a particular niche, and these niches get very specific. We’re not just talking fantasy; we’re not just talking high fantasy; we’re talking authors who specialize in cozy high fantasy specifically with dragons where the main character always dies. Okay, that’s probably a bit of an exaggeration, but it’s not too far off. And it has to be that way because the audience for books and the books themselves are too vast and varied to try to cater to everybody.
And this isn’t just speculation. This is backed up by further data in Bain & Co.’s surveys. When asked what their interests in new video games were in the next year, about two-thirds of gamers said they were most interested in the games they already play (13%), games that are similar (33%), or franchise sequels (19%).
Now, I will say, I don’t fully agree with the findings by Bain & Co. I believe their analysis of some of the data is a bit flawed with hasty recommendations to game company executives.
For one, I think the survey fails to capture the reason many games do so well. While it’s not a safe bet to appeal to too large of an audience, it’s important to acknowledge that the best-selling and sometimes the best-reviewed games are the ones that cater to a particular audience but provide a friendly introduction to a genre for the uninitiated. I believe this was one of the strengths of Clair Obscur: Expedition 33. I’m not much of a JRPG or turn-based combat fan; I’ve only connected with a few of them. Clair Obscur was one of them. The game’s focus on more simple RPG mechanics, characters with far less complexity than others in combat, and narrative served to appeal to fans of the genre while inviting others in to experience the fun without making them feel alienated. In this way, you do want to aim for a generic audience. However, you’re not aiming you whole game; you’re aiming aspects of the experience so that you leave the door open to the rest of the world.
All of the top 10 best-selling games of all time have something about them that makes them easy to get into for just about any person, even if they don’t fall in love with it or play for very long. Games like Tetris, Minecraft, Wii Sports, and Super Mario Bros. hang their hats on simplicity, making themselves approachable for any gamer. Red Dead Redemption 2, The Elder Scrolls V: Skyrim, and The Witcher 3: Wild Hunt embrace compelling narratives and worlds to draw players in who otherwise wouldn’t be interested in the genre. Mario Kart 8, Terraria, PlayerUnkown’s Battlegrounds, and Human: Fall Flat embrace multiplayer as a way to give players a way to interact with their friends. And Grand Theft Auto V… is Grand Theft Auto V.
I also don’t particularly see Bain & Co.’s leap in logic in regards to AI. They asked gamers a simple question: “Thinking broadly, how has your overall comfort with games using AI changed over the past 12 months?” To be fair, the results surprised me. 42% of respondents said that their comfortability with companies using AI has increased either “somewhat” or “significantly.” Another 44% said it stayed the same. With this data, it’s easy to reach the conclusion that Bain & Co. did, which is that “For studios worried that AI adoption carries reputational risk with their player base, this data suggests the window to move is open…”
However, even if other data were to show that this is true, this is a faulty conclusion based on the data collected by Bain & Co. 42% of respondents said that their comfortability with AI increased… increased from what to what? It would’ve been far more valuable to demonstrate this by asking respondents for ratings 0-10 indicating their current comfortability to go along with this. What if the majority of people who said that their comfortability increased gave a comfortability score of less than four? How significantly were those increases, then? It’s the classic fallacy that people fall into when they hear something like “eating xyz food every day doubles your changes of getting such-and-such disease.” That’s scary! I’m never eating xyz food again! Meanwhile, what they don’t tell you is that it increases your changes from 0.25% to 0.5%. If you love xyz food, maybe that’s not enough to stop you from eating it every other day or once a week.
Bain & Co. falls into this fallacy hook, line, and sinker, stating that executives should “Rebuild the focused portfolio on an AI-reset cost base. This is a top-down, organization-wide capability deployment, not simply handing API keys to individual developers and calling it good.“ I think this is a huge mistake. “Top-down, organization-wide capability deployment” is what we’re already seeing. Large companies are several years into doing this, and as someone who works at a large company, it’s painful. Executives fisting AI down their employees throats is what this is advocating for, and I think it’s a short-sighted waste of time and resources.
The reality is that employees will accept just about anything that makes their lives or jobs easier. If there’s an AI that will do that for them, they will find it and use it. Restructuring your entire organization on AI for efficiency and speed actually stands to slow your teams down because they’re forced to do so in order to find a way to incorporate this company-mandated piece of technology into their workflow even though it isn’t helpful. To be fair, that’s more of a hypothesis than a data-driven argument. However, I think it holds as much water as Bain & Co.’s analyses do as they relate to AI.
I think it’s worth noting that Bain & Company has a lot of metaphorical (and possibly literal) stock in AI. In their “About Us” overview, it appears three times in three paragraphs, including the statement: “We’ve led large-scale transformations through every technology wave, and today technology and AI are in everything we do.“ It’s also right there in the first sentence of their “What We Do” page and second sentence of their “What We Believe” page. Needless to say, they have a vested interest in companies using AI, so it makes sense that they would include in their study any kind of reasoning they can that companies should use AI.
Finally, I’ll also note that I don’t like that it’s been very difficult for me to find any information about where their data is coming from. I don’t personally like when these surveys don’t come with descriptions of the sample they surveyed, how that sample was acquired, etc. Even though this likely isn’t often the case, it makes it feel like the data was done in bad faith. For example, Bain & Co. couldn’t even be bothered to tell us how they went about gathering data for their 100 games released in 2023. What was the criteria? Was it random? What were these 100 games? That should be an easy one to include in the report. It doesn’t require any statistical or technical jargon.
Anyway, I don’t take this survey quite that seriously given their practices, but I do think there are some important truths gleamed in it.
AI Game Engine Is Off to Weird Start
The last half of the previous item in today’s post flows nicely into this one. Immens is a company and upcoming game engine that is expected to be AI-first, something every company in every industry wants to be because the executives think it will save them oodles of money while producing the same quality of work, if not better… unfortunately, I’ve seen much to the contrary so far. And what I’ve learned about Immens so far causes my eyes to roll and face to cringe all the same.
The Immens website is sparse, so I’ll be going largely based off of the latest article on the subject from GamesIndustry.biz. To put it plainly, the statements don’t communicate any sort of plan or forethought into how the company will accomplish this feat. However, the general idea is the have an AI-native engine. That is the only concrete details, if you can call that detail.
The article talks to the founders Arjan Brussee, Sjoerd De Jong, and Michal Valient who serve as CEO, Chief Product Officer, and CTO, respectively. What’s immediately noticeable on these three’s LinkedIn profiles is that they have a lot of management experience at a lot of very large companies. To put it in the words of the article, they have worked at “an impressive range of big-name companies, including Unreal maker Epic Games, Guerrilla Games, Boss Key, and Apple…“ To be fair, Brussee does appear to have experience as a Lead Programmer with Epic Games back in the mid-1990s. And Valient admittedly seems to have some knowledge related to computer graphics and other complex, low-level programming techniques. However, little of their available work experiences lead me to think they have the knowledge necessary to build an engine that will rival that of Unreal and Unity.
Supposedly, the company is only the three of them with some “helpers,” so I’m not sure how exactly they’re planning to build this engine. If I were to guess, they’re largely going to use some existing AIs to write the code for them. I don’t see how a team of three core members with seemingly very little modern coding experience is going to be able to stand up such an ambitious project otherwise. That makes me nervous for the eventual quality of the end product. Making an AI-native game engine means having a team particularly skilled in both AI and game engine programming. Each of these disciplines on their own is astonishingly complex, and the two are quite different in terms of knowledge required.
The words themselves that the trio use are even more concerning since they don’t really convey the idea of what they want to accomplish. It makes me wonder if they actually have a vision for what they’re trying to build. Let’s go through some of the quotes:
I’m going to lump a few quotes into this bullet:
“The AI in Immens is about building the engine in a way that is really extensible.” (Valient)
“It needs to be very flexible… You can interchange things. You can append things. You can modify things. You can create your own tools for what you need.” (De Jong)
“You can also create bespoke tools.” (Valient)
Let me be clear. If you’re going to make an engine to go up against Unity, Unreal, and Godot, these are not new, special, or even optional ideas. Universal game engines meant to support the creation of just about any game that a user can think of requires flexibility, modularity, and the ability to extend it. If they hadn’t said these things, they would be considered givens, and if they hadn’t don’t deliver on these, the engine would be considered a failure no matter what.
The article says that, according to Valient, “this isn't AI in the way that many people in the space have become used to talking about the technology. He says that Immens isn't about prompting stuff into existence.“ However, when quoting Valient, it goes on to say, “‘Oh, I want to make this game. I really want my custom particle system. Let me add it there’. So that is one part of how we're doing it with AI in mind. We want to enable people to use AI to extend it faster.” Sure sounds like some form of prompting to me. How else is the user telling the engine what they want? AI works off of prompts. Another way of saying “prompts” is to say “input.” AI typically needs input. While that’s not a universal truth, at the end of the day, if you’re not prompting an AI, then how is this AI-native game engine any different than a normal game engine? Spoiler alert: normal game engines take advantage of AI methodologies under the hood.
“[AI is] really powerful… it can also go off the rails really quickly if you don't pay attention.” To me, this is a great show of how many managers/executives across the world don’t understand AI and/or the unfortunate way that the term “AI” has been co-opted to mean “generative AI.” I’m inclined to believe that it’s both. Believe it or not, a lot of AI up to about five years ago didn’t go off the rails that much. It definitely had failure points or occasional times when AI went “off the rails,” but many of the basic artificial intelligence algorithms were fairly predictable. Consider a chess engine. How much would you say it goes “off the rails”? Maybe to a chess grandmaster in the 1980s they went “off the rails” in terms of playing a slightly inaccurate move that to a chess professional is a huge blunder. It’s generative AI that gets all whacky.
Why? Well, a large part of it is that modern generative AIs are created to try to be good at a great many things when up to this point, AI was built to do extremely narrow tasks and be extremely good at them. Now we have AI that is serviceably good at doing a whole lot of things and really not good at doing tons of other things, but it will still try. AI works best when it has small, defined scopes of use and focuses strongly on that one thing.
“We are building Immens so that when you are using the engine, you are in control of what it is doing… You know what it's doing. You know the history of how it's doing it, so that you know it is repeatable.” (Valient)
…
Umm… what? What does that mean? Again, how is this different from Unity or Unreal? Tasks in Unity and Unreal are repeatable. You can look at the engine’s documentation and know that the engine does the same process, and that process is repeatable. Is the user not in control of what Unity or Unreal is doing? Software largely operates off of human input. You can control what the engine does. Does it always do exactly what you expected? No. But you’re always in control. Again, what is Valient trying to say here?
Valient goes on to talk about the idea of creating a tool that can help you create characters, but you really only need five characters. “Now, you can probably dream it up while drinking coffee and eating breakfast. That tool could be used like five times and then thrown away. This used to be super expensive. You really can make workflows that work exactly for what you want to do.”
Why can’t I do that now by just using an engine or a 3D modeling program along with an AI to help me write the tool? What is Immens doing for me that I can’t already do myself with the current tools available to me?
“You can decide how you want to work, including how much AI you want to use and where you want to use AI.” So is this project an engine, or is it an AI that creates an engine for you? But in that case, you’re not really deciding how much AI you want because you have to use AI at least once to build you an engine that has no AI.
My point in going through these quotes is not to say that I think this project is a bad idea. Rather, it’s to say that it’s hard to say if it’s a bad idea because the founders can’t seem to concisely describe what it is they’re actually building. I’ve formed a few hypotheses:
It’s an engine where each module of the engine (physics, scripting, graphics, etc.) is interacted with via several gen AIs. This is an interesting idea. If each AI can specialize in a particular module, I think each AI can be good enough at one thing for this to actually work soundly. I question what the useability of something like this would be, though.
An engine with a built-in gen AI. That’s it. Not very inspiring, and I don’t think it actually matches a lot of what the founders are describing, but given that I can’t nail down exactly what they’re describing at all, this isn’t out of the realm of possibility.
It could be a generative AI that builds a custom game engine for you based on what you want the game engine to do. After reading the article over several times, I finally came to the conclusion that this is the most likely answer. I also think it’s the worst answer. I simply don’t think it’s going to work. I think generative AI could be decades off from being able to pull off something like this. It’s just not there. Relying on gen AI to program underlying engine logic that is the backbone of any game built with that engine is incredibly risky.
I’d like to end this long item with this quote from Valient: “It has negative connotations right now because it is disrupting everything at a scale that nobody has seen.” I think this is the core problem that a lot of executives have. They don’t understand why people don’t like gen AI in the tech industries. It’s not because they’re just so adverse to progress and are ignorant peasants who don’t know what’s good for them. People don’t like gen AI because it simply does not produce many good results. Its track record is not good. The images it spits out are uncanny and unappealing. The videos aren’t good for anything other than an occasional good laugh or trying to maliciously trick people with a fake political speech. The text it writes feels disingenuous and plastic. The advise and explanations it gives are often flawed at best and provably false at worst. And because of all of these things, people don’t like that other people use it to achieve “efficiency” at the cost of basic quality. It has little to do with it being a new technology and everything to do with how widely it’s been adopted before it’s ready to be.
More GTA VI Leaks
CYBERLEEK, a vigilante hacker group, has leaked gameplay footage of Grand Theft Auto VI, the second time this has happened in the last four years. The first time it happened, the teenager responsible was hunted down and incarcerated. This time is a lot more serious, though. This isn’t a matter of leaked gameplay footage of a very early build. This is leaked footage of a very final-looking build as well as a threat to release the build that’s been acquired to millions. The build does appear to be genuine. Some footage reportedly shows Jason Duval, one of the game’s main characters, spraying bullets into a wall to spell out “LEEK,” something that is mighty impressive to pull off if CYBERLEEK does not in fact have a working build.
Another component of this is the financial aspect by way of a crypto meme coin. CYBERLEEK has watermarked all of the footage with the name of their meme coin with the promise that the hacking group intends to use the proceeds from it to do something unannounced for the gaming community. I think this one aspect of the leaks shows that Rockstar is dealing with more young and stupid people looking to make a quick buck off of scamming others.
If the financial gain is not the group’s main goal, it claims to focus on targeting game publishers that don’t adhere to specific core tenants. I haven’t read these tenants myself in an effort to not give CYBERLEEK the additional traffic nor spoil any aspect of GTA VI for myself, but I imagine they’re largely consumer-based tenants painting the group as a modern Robin Hood for gamers.
Jason Schreier had some interesting points from talking to employees at Rockstar about these leaks. After the first leaks, Rockstar called all employees back to work for five days a week, ending any remote/hybrid work opportunities at the company. Employees obviously didn’t take this well. This was in an effort to reduce the potential for leaks going forward by keeping all work related to GTA VI inaccessible from outside of Rockstar buildings.
With more leaks anyway, though, employees feel understandably crushed. Not only is it heavily disappointing seeing your work plastered on the internet in less-than-ideal circumstances and with a meme coin watermark all over it, but it just serves as evidence that the dismantling of their remote work flexibility was all for naught. I would imagine they’re unlikely to get that privilege back after this either.
If you couldn’t tell, I’m not a fan of this game leaking nonsense. I don’t think it actually does anything to the company long-term. If you look back at The Last of Us Part II’s leaks, they created controversy, but ultimately, it just gave a bigger platform for Naughty Dog and the game. They would spring off of that platform into many Game of the Year awards, including at The Game Awards in 2020. All it does is cause problems and make it more difficult for fans to get their hands on the game they’ve been waiting so patiently for.
Insight Into How Game Trailers Appear at Gamescom
Gamescom’s Opening Night Live event, with none other than Geoff Keighley as the host, has had tons of amazing trailers for already-announced games and games that are being seen for the first time. More information came to light about the process of getting games on Opening Night Live, an opportunity any studio would excitedly jump at given it worked within their marketing plan. Last year, such an opportunity seemed well out of reach when reports suggested that the cost to have a 60-second trailer in ONL rang in at around $119,000. Many smaller games get released on budgets lower than that.
Last August, reports surfaced to the contrary, claiming that it was actually free to have your game’s trailer in the show as long as Keighley liked what he saw.
Naturally, the truth appears to lay somewhere in the middle. According to Keighley earlier this week, there are paid placements and “editorial” placements at ONL: paid placements are essentially advertisements, the time for which the publishers have paid for, and editorial placements are freely given to games that Keighley and some others involved with ONL feel deserve the spot in the show to support the developers.
To me, this sounds very consistent with Keighley’s methodology. I’ve heard several reports on what it takes to get a spot at The Game Awards for your trailer, another event hosted and run by Keighley. It also happens to be one of my favorite gaming events of the year. Depending on the spot in the show that your trailer goes in, a 60-second trailer is estimated to cost up to $450,000. However, smaller developers have told stories of Keighley and his team reaching out to them to ask if they wanted a spot in the show free of charge. Following the Highguard disaster at the 2025 Game Awards show where Highguard was tremendously hyped up by Keighley and showed as the final trailer of the night, the developers, Wildlight Entertainment, revealed that they did not ask for that spot in the show. Instead, Keighley liked it so much that he decided to put it as the final announcement. I’m not sure if Wildlight initially paid to be in the show but then got a higher profile spot or if it was one of those games that got a free ride. Nonetheless, it supports the idea that Keighley and his teams don’t simply chase the money when it comes to trailer placement in events.