5 Video Game Things I Learned | 2026 August 8 - August 14
A long post today! I have lots to say about Saber Interactive and their fight against a former employee regarding AI usage in their upcoming game. I also get into some numbers related to Rixels (the new, more efficient pixel) and Steam festivals. Netflix and GTA VI make it into the post once again as well. Netflix managed to make me furious while GTA VI made me more inquisitive than anything.
The Birth of Rixels
I’ve always lamented that I didn’t get more into graphics programming. I always thought it to be uninteresting. How wrong I was. Especially in the video game space. Some of the most fascinating leaps and bounds in video game technology come from graphics programmers who find ways to accomplish insanely impressive techniques of making games look better while taking up as few computational resources as possible.
4J Studios may have just invented one of the industry’s next leaps. 4J Studios is developing Reforj, an open-world survival game a la Minecraft in gameplay and visual style. Naturally, this means Reforj heavily uses blocky textures similar to Minecraft’s. However, even with what many gamers would call simplistic-looking graphics, “we realised that we were going to run ourselves into a wall with space,” says Chris van der Kuyl, chairman at 4J Studios.
So 4J invented the Rixel (Reforj Pixel) to solve this texture size problem. 4J started off with 512×512 textures, but after chief technology officer Richard Reavy developed the Rixel, textures only need to be 20×20. This reduces the size of a texture by a whopping 98%. Naturally, this would raise eyebrows for anyone familiar with resolution size. When you try to zoom in on a texture, you will eventually reach a point where it loses clarity. However, that’s not true if you’re using vector images. Vector images rely on rendering shapes as opposed to individual pixels. While vector graphic displays have fallen out of use in favor of raster displays, vector graphical techniques are widely used to reduce the reduction in quality for images as they are resized.
Because 4J’s Rixels use vector shapes, not only is the new 20×20 texture much, much smaller, it actually looks better because it doesn’t lose visual clarity as it scales up. This allows Reforj to maintain more consistent performance and do more with less. Rescaling small textures is much easier for computers than loading large textures. In the case of a 512×512 texture vs. a 20×20 texture, the sizes are 4 MB in memory and 72 KB respectively (the 98% reduction). Let’s say we have 2 GB of memory allocated for textures. With 512×512, you can have 500 individual textures. That’s a lot of textures! That means you have enough space to store nearly half of all of the block types in Minecraft. And since you can reuse these textures, that’s way more than enough textures to render whatever you want to the player in terms of a landscape. Now let’s take 20×20… you can have over 27,700 individual textures in the same amount of memory. We’re not just talking about freedom from a development perspective. We’re talking about a g
While I’m not an expert in graphics, I feel that this could be a revolutionary technique that is built upon over the years to create new industry standards. In its current implementation, its usefulness is likely limited, though. “It may matter less to photorealistic games than it does to a stylised game,” says van der Kuyl. However, van der Kuyl shares my optimism that this could be built upon. “It'd be interesting to see what people would do with it in other contexts.” He says that the company would be very interested in the idea of licensing the technology at some point when the patenting process is navigated. I think that’s the best way to grow the technology. Allowing others to use it creatively in the wild will truly unluck what the Rixel can do.
Netflix Kills Night School Studio
I. Knew. It. I’m so devastated by this news today. Netflix has made an enemy today… well… an enemy that still pays for a Netflix subscription… But an enemy nonetheless!
In all seriousness, I saw this coming from a mile away. And by a mile, I mean six weeks away. Six weeks ago is when Netflix announced a new game exclusive to their cloud gaming platform called Unhinged. It was developed by the talented team at Night School Studio, the developer behind Oxenfree and its sequel, as well as Afterparty. Over the past 10 years, they’ve become synonymous with high-quality narrative games. Now, Netflix has shuttered the studio and at the same time touted the numbers from Unhinged. Back in July, co-CEO Greg Peters called it one of the company’s “2 most successful cloud game debuts, really solid numbers that put it in the top tier of game performance for us.”
I’m going to commit the cardinal sin and quote myself from June 27, 2026, when I talked about the announcement of Unhinged:
“I also wonder what the metric for success will be for Night School. There certainly will be a metric they have to hit in order to keep the studio off of the chopping block. However, the game has no associated purchasing or ads, according to Netflix. So how will Netflix determine the success of Unhinged? Even large corporations that are fully engrossed in video games don’t seem to know how to set reasonable expectations for their first-party studios to hit. What does a company primarily associated with the film industry think a successful game looks like on Netflix?”
It turns out that entertaining millions (according to the game’s director, Sam Warner, on LinkedIn) and setting the bar in terms of Netflix’s cloud gaming doesn’t cut it, so Netflix is cutting the studio.
Unfortunately, in retrospect, it seems like Night School Studio never had a chance, and I think it’s because Netflix cloud gaming as it exists today doesn’t have a chance. This news from this week has inspired me to write a separate post about how Netflix’s gaming initiative is doomed.
In additional to Night School Studio, Netflix is dropping Moonloot, a studio that never even released a game. With both of these companies now gone, Netflix has managed to start or acquire seven different studios and cease operations with five of them in the span of just five years. Other victims include: Team Blue, Boss Fight Entertainment, and Spry Fox (Spry Fox was sold back to its founders, so the studio wasn’t shut down).
Netflix also made cuts to their internal game studio staff.
People Want to Pay More for GTA VI
After last week’s announcement from Strauss Zelnick, CEO of Take-Two Interactive, that Grand Theft Auto VI preorders were in a better spot than the company even expected, Sensor Tower, a digital intelligence firm, has indicated that roughly 89% of those preorders have been for the $100 ultimate edition.
Color me surprised, particularly after the backlash from gamers about the content that was being hidden behind the ultimate edition. When Rockstar announced that the base price of the game would be $80, gamers largely sighed in relief that the base game wouldn’t be $100 like some analysts speculated it could be. However, in the same announcement, we found out that some key cosmetic customization options are being locked behind the ultimate edition’s price tag. Many Grand Theft Auto fans didn’t take kindly to this. They took to the internet to complain about how integral the customization mechanics are to Rockstar games in general, including the GTA series. Historically, gamers don’t like the feeling of being nickel and dimed, and many believe that’s what’s happening here. Clearly, that was either a vocal minority or the people complaining about it are so invested in the character and vehicle customization that they felt like they just had to spend an extra $20 to get the full experience.
As senior vice president and general manager of Sensor Tower Video Game Insights Karl Kontus points out, it’s common for premium edition sales to take over during the preorder period. This is because preorders and premium editions tend to target the same type of player: the uber fans. However, 89% is crazy. Kontus says, “A 50% premium share during preorders is a great number. Typical premium shares for games tend to be at around 10–20%.“ In terms of translation to total sales, Kontus estimates that GTA VI will land somewhere around 25 million preorders. For reference, that’s how many copies Grand Theft Auto IV has reportedly sold in its 18 years on the market. GTA VI is poised to hit that before the game even releases. Another interesting point: if GTA VI releases with 25 million or more preorders, that may be the first ever instance of a video game becoming a top-50 best-selling game of all time before it even hits store shelves.
Saber Interactive and Former Employee Fight About AI
A war has begun between Saber Interactive—the developer/publisher behind World War Z and co-developer behind remasters for games like Crysis and Tomb Raider—and a group of the gaming community headed by former Saber Interactive employee Stella Sacco. Following Saber Interactive posting the trailer for the upcoming simulation game Rideshare "Stimulator", Sacco took to the internet to let everyone know that “I was lead writer on this one! And Saber replaced me with ChatGPT midway through development. All the passenger voices were AI too.” She also pointed out that the game’s Steam page had no generative AI disclosure, something that likely indicates that it was not revealed in Saber’s completion of the Steam Content Survey.
Sacco also claims that the decision to use AI content in Rideshare “Stimulator” was pushed largely by a particular “c-level“ executive whom she later seemingly revealed to be Matt Karch, Saber Interactive’s CEO. Her account spells out a firm stance by Stacco that “players hated AI in Dec. 2023 and they’d hate it even more by the time the game came out,” but “He didn’t listen to me, he didn’t listen to my boss, he seemingly didn’t listen to anyone at all.” Rather, Karch allegedly “insisted ‘gamers love new tech’ so using AI would be ‘great marketing’.”
Over the past several days, there has been back-and-forth between Sacco and Saber Interactive in the media, and Sacco’s clarification of Karch as the generative AI-loving suspect came as part of that. Specifically, it came in response to a statement from Saber in which the company denied Sacco’s claims. “Neither Saber nor Unigine have replaced any writers with AI for Rideshare or any other game," the statement says. It goes on: "Saber's staff of writers has tripled since 2023. Rideshare's story was written entirely by real people.” The response also clarifies that game will in fact feature generative AI “as the number of passengers in the game is infinite. We are firm believers in using AI tools to improve or enhance gameplay experiences and improve NPC behavior as we have done since Saber was founded in 2001.“
Following the Sacco retort, Karch went on the offensive without hiding behind any company letterhead. The following is the entirety of Karch’s statement:
“Stella who? I had to ask Claude because I have never spoken to her or seen her. We have already said we are using AI for certain parts of the game and will update the Steam page once we have finalized the feature set.
“Let me clarify that the entire story script, were it to win an award, would go to someone who can walk up on stage and receive it. It was always our intention to reveal this until Stella violated her confidentiality clause and revealed this before we were ready.
“We look at our use of AI in this particular game, which will be a lower priced title, as a plus for consumers.
“As far as Stella is concerned, she was replaced by someone more talented and terminated for a variety of reasons a few months later. From what I can gather she enjoys controversy even more than I do. I would frankly in retrospect have been happy to replace her with AI. At least we would be dealing with someone programmed to be honest.”
Sacco has mostly taken a step back from responding further, but did leave off with one post that states: “I think it should be obvious to anyone who reads that statement that he’s lashing out because he did not want people to know there was AI use in the game.“
Now, the reason I posted Karch’s statement in its entirety as opposed to cherry picking is because I sort of want to go through the statement point by point:
He starts off in a less-than-dignified way by asking “Stella who? I had to ask Claude because I have never spoken to her or seen her.” In my personal opinion, these two sentences wreak of someone too far up on the corporate ladder to care about his employees (current or former) and so lost in large language models that he doesn’t take any AI criticism seriously. Furthermore, while his statement comes off jokingly, it does make me wonder if he did in fact ask Claude. And if he did ask Claude, did he just ask the basic Claude model, “Who is Stella Sacco,” or has he/Saber Interactive fed employees’ information into Claude to query against the data, allowing Claude to give specific information about Sacco’s employment with the company?
Karch takes deliberate action in accusing Sacco of violating her confidentiality agreement with the company. To me, whether or not Saber could legally go after Sacco for this, I don’t think this holds up in the court of public opinion. By saying that “It was always our intention to reveal” the usage of generative AI in Rideshare “Stimulator” when “we were ready,” the Saber CEO is directly admitting to actively hiding this practice despite Steam’s policy requiring its disclosure in its Content Survey. Said disclosure would presumably then show on the game’s Steam page. I don’t necessarily have data to back this up, but I would venture to guess that most of the gaming community does not see generative AI as a feature in their games that the company can feel free to cover up the details of so as not to spoil the game.
Saber Interactive’s initial response to all of this stated that “the Steam page will be updated once the feature set is finalized,” supporting Karch’s claim that it would be revealed eventually. However, I personally believe this to be either tremendously incompetent or complete bologna. As of today, the AI disclosure now appears on the Steam page. Either it just so happens that the feature set was just now complete, coinciding with these words from Sacco, or someone at Saber Interactive realized that they lied on their Steam Content Survey and rectified the situation. The former sounds unconvincing, to say the least, and the latter shows poor video game business acumen.It’s very strange to me that Karch appears to be coping with his company’s decision to use AI-generated content in their game by specifying that it “will be a lower priced title.” Does it matter if it’s a lower price title? What does this statement actually mean? The only thing I can think of is that he’s using some kind of twisted logic to imply that a lower price means that the game has a smaller budget which means the team is smaller, and it’s more understandable and expected for a smaller team to leverage generative AI in its development. I don’t believe the dots all connect there. I think it’s far more likely that he’s trying to mitigate outrage by devaluing his company’s own product, implying that people shouldn’t be so upset about it since they won’t pay AAA prices for the game. A very sad tactic, and regardless, a sentence that makes this whole statement feel hastily thrown together without the PR department’s knowledge.
The final paragraph is a foolish one with a tremendous lack of taste that Stella Sacco describes as “truly horrendous.” Regardless of what the truth of this all is, Karch continues to paint himself and his company in a terrible light. To start your statement by saying that you have no idea who Stella Sacco is and then conclude it by making qualitative judgements on the work that she can do being at best comparable to that of an artificial intelligence is laughably contradictory. So much so that it further cements this statement as a flimsy one written by an angry executive with no knowledge of the situation or tact for public relations.
Now, despite my obvious disgust with Karch and Saber Interactive in their handling of this, I think it’s worth noting that we have little evidence outside of word slinging. Usually the truth of the matter ends up lying somewhere in the middle of two heavily opposed sides, and I think that’s the potential case here.
Sacco talks about AI-generated writing and voices being employed by Saber for Rideshare “Stimulator”. In my opinion, they have a very specific use case that warrants the use of such things. If you want to have a truly infinite gameplay experience that doesn’t get stale, that’s a place where AI has tremendous potential in the video game space. In the case of a ridesharing simulator, the world feels much more alive and seamless if dialogue, characters, routes, and voices don’t repeat. The game feels much more full this way. Obviously, doing such a thing is impossible without AI; you can’t manually record infinite voice lines, create infinite character combinations, etc.
I think it’s also important to note that Sacco was let go from Saber Interactive more than two years ago according to her LinkedIn profile. The game would’ve been in a very different spot in development. AI-generated writing and voices are actually an integral part of the game development process for many developers, and it’s been that way for quite some time now. These works help create placeholder content for the team to work around and test against without going through the exhaustive and expensive process of producing final dialogue and voice lines beforehand.
With that said, I think the most likely reality here is:
Matt Karch is just another business guy drooling way too much over generative AI. He may have been pressuring the organization to use generative AI more, rewarding those who did, and and essentially punishing those who didn’t.
Nonetheless, Saber found a legitimate use case for it in Rideshare “Stimulator”, so they began the implementation of generative AI for specific features to support infinite content in the game while keeping the main narratives human-written.
Saber Interactive knew that this functionality was key to the defining features of the game they were making, but they also knew that the public would not be happy about it. So they did the best they could to keep it out of the public’s view. I personally believe despite their statements, they were likely never going to say anything unless the AI in the game proved to be a huge success with players either not noticing or believing it to be a genius use of AI.
In July of 2024, Stella Sacco was terminated from Saber Interactive due to very boring reasons: they didn’t like the quality of her work, her style didn’t gel with what they were trying to do, she was missing deadlines, she had HR violations (such as being late consistently, having no-call-no-shows, or something like that), or some other reason. However, around the same time, she and her boss were fighting tooth and nail against the idea of using generative AI in their game. When she was given the boot, it seemed pretty obvious that they would just replace her role with AI-generated writing, given the context at the time.
I’m making three enemies here: enemies of those in support of Sacco (including Sacco herself) for implying that she’s mistaken or being less than truthful, enemies of those in support of Saber Interactive (including Karch) for criticizing their handling of this whole debacle from the very start of development of Rideshare, and of you reading this because it’s so damn long. Thanks for reading!
Steam Festival Volume
I always keep up with what Chris Zukowski is doing over on HowToMarketAGame.com. This week, he threw up a blog post about third-party festivals on Steam and how they can increase visibility. The post is short and sweet. It looks at how many festivals there have been since the feature became a thing on Steam in late 2019 and how many games were in them. According to Zukowski, about 31% of all Steam games released in that timeframe have participated in at least one festival. That’s higher than I would’ve thought.
Now, in order to get into these third-party festivals, developers and publishers have to apply to them and get accepted, so how many festivals a game participates in is highly dependent on its appeal to festival organizers. Zukowski shows a graph and corresponding table that puts the median game at just two festivals. The upper 30% make it into three, 10% make it into six, 5% make it into 10, and the top 25 games make it into over 30.
Initially, I figured this was likely because the top 25 was comprised of games that have been out for at least a few years and have had the chance to make it into more festivals. How wrong I was! In fact, 13 of the top 25 were released in either 2025 or 2026. And an additional seven haven’t even been released yet. That leaves just five of the top 25 releasing before 2025.
New theory: newer games are actually the ones that have had more opportunities! When a game releases, I imagine the publisher’s focus gradually shifts off of that game and toward upcoming games to try to capitalize on those releases’ prime times for sales. So the games that just recently released or are set to release in the future have had their full marketing timeline to take advantage of this feature, as opposed to games from 2019 and the early 2020s that didn’t get that chance since the festival feature didn’t exist and fewer festivals existed.
Zukowski points out that certain types of games appear more likely to land higher numbers of festivals. First of all, the top 25 is crowded with cozy and narrative games. This is attributable to more festivals for those types of games. The vast majority of festivals are for all genres, but by far the next two types of festivals are “Cozy” and “Narrative” ones. There’s then the advantage of having a publisher. Games without publishers have less manpower, less money, and less time, so it’s harder for them to mass apply to festivals throughout the year to get into a bunch. A publisher provides the resources necessary to get your game seen. This includes connections with other publishers since a lot of festivals are put on by publishers themselves.
I thought this was an interesting tidbit. I really enjoy learning about Steam and how marketing works on the platform. I love how much data is available and how Valve handles the storefront.