The Unreal 6 Reveal is Epic's DLSS 5 Moment | 02/07/26
Integrating Generative AI into workflows is the least of your problems.
Hello one and all and welcome to this week’s AI and Games. This week we’re back to talking about the industry and all that chaos that’s going on, with a particular focus on Unreal Engine 6.
Follow AI and Games on: BlueSky | YouTube | LinkedIn | TikTok
Announcements
Obligatory round of announcements for this week. Lots of stuff happening and coming your way!
AI and Games Events
Your weekly reminder of the events we currently have in the calendar!
Gamescom Dev AI Tracks [24th-25th August]
Tickets are on sale now, with our AI tracks part of your regular Gamescom Dev ticket, complete with access to the Gamescom trade show.
Sponsorship opportunities for the AI tracks and across Gamescom Dev are also available.
AI and Games Conference 2026 [10th-11th November]
Early bird tickets will be on sale until end of July.
Call for speaker submissions is also open until August 2nd.
Accepted speakers will receive a free ticket.
All other submissions will receive voucher to purchase ticket at early bird rate.
Plus another week, another sponsor, we’re excited to welcome Studio Atelico on board as our Scholarship Sponsor for 2026. Thanks to their support, this helps us ensure that even after early-bird sales close, we will be keeping ticket costs for students under £100, while indie tickets will also be subsidised. As you can imagine these events are expensive propositions to run, but we still want them to be accessible to as wide a range of developers as we can.
Sponsorship opportunities are still available, and feel free to reach out to us for more info, or visit the conference website.
New Case Study: Forza Horizon 6
We released a new case study episode of AI and Games earlier this week. With the recent release of Forza Horizon 6 I figured why not revisit a topic that has proven quite popular over the years. The big reason for this was the discourse happening online surrounding ‘Bowie Knife99’: one of the built-in AI characters whose behaviour proved so chaotic at launch that they became something of a meme within the Forza community.
So in this latest deep-dive we explore how the Drivatar system has changed in recent years, but also what are the circumstances that led to a bot like Bowie Knife99 arising in the first place. The answer is, unsurprisingly, rather complicated.
Podcast Kicks Off Next Week!
Last week we announced that the AI and Games Podcast is launching this summer! We’re currently putting the finishing touches to the first episode, which is a follow-up to our 2024 overview on AI in the video games industry.
Plus we’re slowly rolling out the feeds so you can subscribe on your chosen platform. You can now find the AI and Games Podcast on:
Spotify | Apple Podcasts | YouTube Music | YouTube | PocketCasts
Be sure to check out last week’s issue with a deep-dive into what we have planned with a bunch of guests lined up in the coming months.
Epic Games’s Pulls an Nvidia with Unreal Engine 6
In an effort to try and wrestle back our scheduling to having the newsletter launch on Wednesday’s - I like having it midweek don’t you? It breaks my week up nicely - I was torn over which groan-inducing headline I’ve had the pleasure of reading about in recent weeks that would receive the full force of my ire this week. Despite Sony, Microsoft, Epic Games, Meta, and Valve all having news stories that flirted with the dreaded AI hype conversation, it was Epic that officially took the biscuit. Besides, I started writing this issue on Monday 29th, and I figured Xbox would drop a whole new round of god awful headlines as their end of fiscal wraps on June 30th.
Spoiler alert: Boy is this going to be a tough couple of weeks to report on.
So I figured why not indulge and have a good rant about Epic Games. Well, no not necessarily about the devs at Epic. I think they do a lot of great work. Rather this is about leadership at Epic Games, and how Tim Sweeney and those around him are successfully crossed the Rubicon, in promoting their efforts for Unreal Engine 6 that I do feel will prove to be akin to Nvidia’s hubris with DLSS 5.
Unreal Engine’s Roadmap Revealed
Epic Games hosted Unreal Fest in Chicago from June 16th - 18th and in amongst a whole heap of presentations was details about the next iteration of the Unreal Engine. After we received a tease of the new engine courtesy of the Rocket League Championships in May - which was hilarious to me given Rocket League was still in Unreal Engine 3. The big announcement introduced a number of new and updated features as we transition from UE5 into Unreal Engine 6:
UE6 = UE5 + UEFN: Unreal Engine 6 is seeking to merge the two currently disparate development tools that Epic Games have built. Notably the Unreal Engine for Fortnite (UEFN) editor. The idea being that any game that can be built in UE6 can be released standalone, or deployed within Fortnite.
Portable Content: A big push for interoperable content across games. Allowing for assets and other content and code to be able to transition from one project to another seamlessly. So for example, a Fortnite skin can then more readily transition to another game project.
New Scene Graph Framework: Given the push to merge UE5 and UEFN, this means that the existing ways in which game scenes/levels are modelled in Unreal Engine is being deprecated and removed. No more ‘actors’ for representing objects, and the use of Blueprints for visual scripting. This is all being migrated to the Scene Graph system used in UEFN.
Native LLM Integration: The big conversation point for many was that starting in UE5.8 there is now a plugin to host an MCP server - a Model Context Protocol server - that allows for you to connect Large Language Models such as Claude, GPT or Gemini into the engine to support design and development of environments. Ranging from placing of library assets to integrating with the existing procedural systems and lighting and modifying values on actors and materials as it goes.
The Verse Programming Language: Unreal Engine 6 will be transitioning the underlying programming systems from C++ to Verse, which is the language currently used in UEFN and is described as “the programming language of the Metaverse”.
Now the funny thing is, I was busy working when this all went down and couldn’t follow along as it happened. However the reaction across social media has been in outcry over the AI models - which is something Epic have been working on for a while and as I’ll discuss what most developers won’t really use. However, I feel like all the noise on social media and in the gaming press has fundamentally missed the mark.
This whole thing is, in my opinion, an absolute nightmare. For me the issues are numerous, and the last concern I have is with the LLM integration
The desire to merge UEFN with UE5, which caters less to the users of Unreal Engine, and more to Epic Games push for metaverses and NFT-like content creation pipelines.
Removal of Blueprints is a catastrophic situation for studios that rely on visual scripting tools, and the Scene Graph does not cut it.
The push to transition from C++ to Verse isn’t just a problem for games studios that use Unreal, but the games industry at large, and the academic environment that supports it.
Oh yeah, that LLM integration means very little in a world where token costs are going to continue to increase.
It speaks to a fundamental assumption that Epic knows better than the industry that surrounds it, and I have yet to speak to anyone I know across the industry who was excited about any of these announcements. It feels like the DLSS5 announcement from Nvidia all over again, where a bunch of ideas are presented that developers in general aren’t jazzed about, and that’s because the provider is running on an assumption that they know best.
The LLM Integration
Now the introduction of the LLM MCP plugin is, as I say, the least of my concerns, and frankly it does what I expected it to do in a way that Unity has pushed for developers to embrace in recent years. Plus we had Gabriel Ware and Juan David Hincapie from Google present at the AI and Games Conference 2025 about their experiences of integrating Gemini into Unreal Engine 5 in ways that will no doubt have influenced this.
While I understand people’s hesitance and resistance to this, I also see benefits from trying to integrate LLMs into workflows to help with specific aspects of their work. Plus given it works directly in-engine, it allows for the opportunity for users to iterate on ideas in the editor, and then pivot to the MCP server call if necessary.
You can see the LLM MCP plugin in action in the video above, and while it’s interesting for sure, there’s a lot of issues hidden courtesy of an obfuscated presentation and corners being cut and questions not being addressed that people need to be wary of:
It’s a Plugin: The MCP server is currently am experimental plugin, and is not part of the engine by default.
Time Taken: Note that this video speeds up many of the production times it needs for these generations. The apartment example takes 3-4 minutes to generate each time. The skybox iteration example takes close to 10 minutes. How useful will this be in practice when I’m waiting 5 minutes at a time for scene changes?
Token Costs: Many of these processes are costing thousands of tokens per call, given the system has to iterate internally on the prompts raised by the developer, and how that cross references against increasing token costs, and what budgets studios can allocate to this is nicely avoided.
Efficacy for Complex Prompts: A lot of the prompts given here are relatively simple, but have cascading effects - and the big changes in the scenes seem to be driven more by Unreal’s procedural systems than the LLM. Sure let’s reconfigure lighting, manipulate materials etc. But what about larger and more complex changes? I doubt this would work as intended and requires users to work more gradually.
Context Management: In the demo calls you see a lot of behaviour iterating internally, and it suggests there might be multiple agents running to generate this output? If so, how do they guarantee context is retained over time?
Mult-User Interoperability: How can multiple users be working in the engine at once and what are the implications of having multiple users interacting in this workflow?
Frankly I never take tech demos like this at face value, it’s already obvious that they cut corners and speed parts of it up to show its impact - I mean that video above would be around 90 minutes long if they didn’t speed up the demos. Plus how Epic present this feature versus how people work in their engine are always very different things. So it’ll be more interesting to see what happens when studios are able to show how or if they found this useful and why.
The Bigger Issues
As mentioned, there are much bigger issues laced throughout this, and it speaks to a desire for Epic Games to effectively fold Unreal Engine into the Unreal Engine Fortnite Editor. In truth it reads much more about Epic’s priorities towards their ambitions with treating Fortnite as their own AAA Roblox platform, rather than putting any thought or consideration into how, where, and why, developers use the Unreal Engine.
It’s worth stating first and foremost the Unreal Engine is a big deal in the premium PC/console video games industry. It’s used by an increasing number of studios around the world or has an influence on how studios operate their own engines or integrate developers into their own ecosystem. It’s important to remember that when Unreal Engine began to become more public courtesy of Unreal Engine 3 and Unreal Development Kit (the education edition of UE3) in the mid-2000’s followed by Unreal Engine 4 in the early 2010’s. Sure, not every studio uses Unreal but it has helped shape a lot of how people perceive and engage with engines, and allowing for it to be more accessible means that it has had an impact on the expectations studios have of developer skill. It’s not uncommon for a studio that still runs their own proprietary tools to expect people to have working knowledge of Unreal given it’s ‘closer’ to how they operate when compared to other engines.
Unreal Engine since its inception has been C++ based, and it helps create a conduit of talent. Many studios rely on programmers who can develop in C++ either because of Unreal Engine or their own proprietary engine is built in that language. Sure the likes of Unity and Godot also help proliferate C# as a programming language, but my point being that Unreal has been instrumental in ensuring corners of the industry have knowledge of the programming language, and critically a lot of Higher Education is built to support that as well.
Much the same can be said for visual scripting tools like Blueprints. Not only does this help provide for designers and other technically proficient developers to work within the game engine, but also it’s built such that entire games can be shipped pretty much with little to no coding. A recent example of this is the award-winning Clair Obscure: Expedition 33 which was built almost entirely in Unreal Engine Blueprints, with only around 5% of the game implemented as code.
So to threaten to remove both of these - and notably without much indication on what alternatives there will be to Blueprints strikes me as ridiculous. Transitioning from C++ - a well-known, highly-regarded, regularly iterated up and standardised programming language - to Epic’s own scripting tool largely unknown outside of the UEFN community is ridiculous. I mean sure ‘Unreal C++’ is technically its own beast (notably with a macro-based reflection and metadata exposure functionality, plus reduced memory management), and that creates a lot of confusion and frustration at times, but again there is an established ecosystem where people can learn C++, and then learn how it works in Unreal. The new approach is to work primarily in Verse: a programming language that leans heavily on ideas from functional and logic-based languages (which most aspiring engineers don’t learn) and is built entirely by Epic Games themselves. Sure it has some similarities with say Python, but this is now getting us into the weeds of obfuscating memory management in ways I don’t like the sound of. I mean that’s just hubris at its finest, and sadly some of us still remember how god awful Unreal Script was in practice.
I mean from what I’ve played with it Verse is generally fine as a programming language, but it raises a lot of questions:
Why the push to move towards your own proprietary language?
They state that C++ will still exist in UE6, but what breakdown there is between the two is not clear.
Why do I have to use this when it is going to run via a LLVM compiler that will transcode it into LLVM IR before it then executes?
As I say it sounds like C++ will continue to be a big part of Unreal Engine 6 going forward, but right now none of this sounds particularly appealing for studios with existing expertise and knowledge about the engine. Switching languages, tools, and even entire paradigms takes time. I find it amusing that in the same day when they announce how LLMs are going to help optimise your workflow in Unreal, they also reveal a set of changes that would forcibly delay any production by months should they be foolish enough to migrate to UE6 during development (and no, nobody is going to do that - we’re not mad). With entire production workflows being up-ended by a new feature that doesn’t feel ironed out yet. Meanwhile you have university courses that are built around training students on how to use Unreal Engine - heck their staff get certification for it. Universities can’t just pivot entire degrees on a dime; these are slow-moving beasts that take time to update and address the correct needs of their students.
Perhaps they will address my concerns over time. But for now, I wouldn’t be surprised if a significant proportion of the games industry just keeps their local installs sitting at around Unreal Engine 5.7 or 5.8 as they wait to see where the dust settles. Right now this all seems to pushed towards supporting studios that work in Fortnite, or aspire to make games like Fortnite, or in Fortnite - again Tim Sweeney’s long-term ambition of turning the game into his own personal Roblox. None of these changes strike me as addressing a known problem for developers themselves, or catering to their needs. Rather it’s all about servicing Fortnite, or building games in Fortnite, which represents but a small fraction of the adoption and use cases the engine has. Streamlining with UEFN and creating workflows that cater to this broader ‘Robloxification’ of their platform does nothing but to line the pockets of Epic’s executives. At a time when Fortnite’s numbers aren’t achieving that which allows for Epic to sustain themselves,- which don’t forget led to 1000 developers being laid off but a couple of months ago - I fear they’re throwing Unreal Engine under the guillotine in servicing Fortnite’s needs.
It’s important to remember that despite all the boom and bluster surrounding Fortnite, much of Epic’s longevity has come from providing an engine that studios want to use, and the ecosystem that creates around it. Failing to respect that, or working in ways that contradict it, could prove disastrous for them in the long run.
Wrapping Up
As we wrap up, I wanted to give a big shout-out to fellow AI and Games Conference organiser Matthias Siemonsmeier and his team who have been hard at work on building a new wildlife system into the popular survival game SCUM that allows for all sorts of animals to be found wandering around in the open world for you to hunt or evade! You can see Matthias talking about this new system in the dev diary video below.
I know Matthias has worked very hard on this in recent weeks - he’s been rather sleepy in recent conference meetings - but was driven largely by a passion for making this work and delivering on a feature for a game I know he loves to play. Plus keep an eye out for one of our very own AI and Games plushies hiding in the background.
It looks like a cool new system, and lots of hard work went into it. Maybe we can convince him to talk about it via the various avenues of dissemination we have around here huh?
See you next week!
There might be a misunderstanding regarding C++ and Verse in UE6. From what I read, it's that Verse replaces the scripting layer currently provided by Blueprints. The core engine will still be C++ like before, that's not being replaced by Verse. That's what I took from it at least.