Iakov Sumygin

UE5 game dev is chaotic—your workflow doesn’t have to be. In this 3-part series, learn how real teams use Perforce, CI/CD, and C++ best practices to keep projects bulletproof. Learn how to: ✅ Set solid project foundations ✅ Master Perforce workflows ✅ Build game-ready CI/CD pipelines ✅ Write efficient C++ under pressure 📅 Sep 30, Oct 2 & Oct 3 — 9 AM -10 AM CDT REGISTER NOW: https://lnkd.in/gdmCu5ET #VisualAssist #Assembla #UE5 #cpp #developers #coding

2

A couple days ago I gave Claude Code the ability to debug GPUs with RenderDoc. Now I taught it Apple's Metal ecosystem.                                                I built a second skill — metal-ai-skill — that connects Claude Code to Apple's GPU toolchain: xctrace, Metal validation layers, shader compilation, .gputrace frame capture, and screenshot-based visual  inspection. When I describe a rendering bug on macOS, Claude can:  - Screenshot the app and actually see what's wrong (multimodal vision)  - Capture a .gputrace frame and inventory every buffer, texture, and shader  - Compile shaders with -Weverything to catch static issues  - Read source code to trace data flow from CPU to GPU  - Fix bugs and re-screenshot to verify — in a loop until it's correct  The interesting challenge: Metal's CLI tooling doesn't give you buffer data from programmatic captures (unlike RenderDoc's rdc-cli which gives you everything). So I built an autonomous debugging workflow  — Claude gathers signal from multiple sources in parallel, checks what data is actually available, and falls back to source code analysis when GPU data isn't there. Every path leads to a diagnosis.  In the demo, I give Claude a Metal cube renderer with 4 intentional bugs: collapsed geometry, flipped Y axis, swapped color channels, and zeroed alpha. Claude screenshots the broken output, reads the  shaders and Swift code, diagnoses all 4 bugs, fixes them, rebuilds, and verifies the output — all autonomously.  The RenderDoc skill taught me the pattern. The Metal skill taught me the fallbacks. The real insight: when you can't inspect GPU state directly, combine enough indirect sources and you still get there.  Open source (MIT).  GitHub: https://lnkd.in/g8n8qb6J  Video: https://lnkd.in/gryAY4YR  #ClaudeCode #GPU #Metal #Apple #GraphicsProgramming #AI #GameDev #OpenSource

114

11 Comments

Following up on my previous post about using Modular Rig for simple environment animation, but this time focusing on a mechanical prop. This prop uses slightly more complex module than the previous one, and the results were honestly a bit surprising (in a good way). ➤ What worked well ✅ Moving beyond single-joint motion is doable, as long as the prop parts are spitted with animation in mind. ✅ Building a skeleton and painting skin weights directly in Unreal works well for this prop, which has a moderate complexity. ✅ Mechanical modules like Hinge and Piston are already well supported out of the box and easy to reason about. ➤ Where friction shows up ⚠️ This still isn’t a full DCC replacement. The quality of the result depends heavily on how well the asset is designed and segmented upfront. ➤ Bigger takeaway For me, this reinforced that Modular Rig lowers the barrier to keeping work entirely in-engine and expands what environment art and level design can include, especially for smaller projects where DCC round-trips aren’t available or are too costly. On top of that, it could potentially enable iteration to happen directly in the level. Still exploring, but I’m curious how others are thinking about where this kind of workflow fits into their pipelines. 🤔 #UnrealEngine #UE5 #TechnicalAnimation #GameDev #TechnicalArt #LevelDesign #ControlRig #AnimationPipeline

353

13 Comments

Just found this undocumented feature added in UE5.5+ which allows standard Editor Utility Widgets to be dockable as a main tab instead of requiring it to be nested. This is especially helpful for those making widgets that require larger real estate or embedded viewports. I was digging in C++ source the last little while to see how to achieve this but looks like it was added in a different, and more accessible form, for newer versions of the engine already.

20

What a great #gamescom2025 for Gamebreaking Studios! Taylor Hellam and I had more than 40 meetings covering co-development and pitching Dungeon Rampage, and we gained significant traction with some great potential partners. I am continuing with follow-ups this week, but I wanted to take a moment to share some of my key takeaways 📌 from this year's conference (Taylor's shared his as well, so please also check that out!): 📌 Games are Still in Motion, and Co-Dev is 'Hip' As much as we are still feeling some of the pain points of over-estimation of growth in our industry, there are still games being made albeit with a different core production strategy. We are seeing smaller, more focused in-house teams with a higher input of co-development partners owning larger mandates/features. This is an exciting time for studios who have industry-leading talent looking to bootstrap their own growth through co-development opportunities with some very interesting upcoming projects. As a result, XDS this year is shaping up to be critically important and Gamebreaking will have boots on the ground to continue some of the important conversations kicked off last week. 📌 The A, AA Void We are observing a very visceral aversion to risk right now in raising for games in the A, AA space. There seems to be a gap forming where the only games getting greenlit are either major AAA, AAAA franchises or small (<1m) indie titles. What seems to be occupying the A, AA space is exclusively re-masters of back-catalog games or games on established IP/Licenses. It's an interesting time for studios in the A, AA space who may have to lean harder into co-development to bridge this gap until funding returns to this part of the market for new IPs and creative ideas. 📌 Preparation and Confidence are Key When pitching, it's very important to understand each publisher/VC can be very subjective and it is critically important to have the right supporting materials for each case. Taylor did an amazing job prepping multiple decks for us to pull from. These ranged from covering more of the narrative of our studio and the development of Dungeon Rampage, to very specific statistics decks running out the best, average, and worse case revenues. A huge thank you goes out to Jacques Benchetrit and Garry Seto for their impactful feedback at the beginning of the week, which allowed us to refine our materials and give us the confidence that our numbers made sense. This confidence translated directly into our pitches, where the publishers we met with were very impressed with how prepared we were for each meeting. 📌 Final Takeaway, cut in some vegetables and water/electrolytes at regular intervals. Gaffel Kolsch 🍺 and Crispy Pork Knuckle 🍖 are amazing, but not something to eat at every meal 😅 Thanks to everyone for a great conference!!

31

2 Comments

Vulkan vs. OpenGL — Why the Magic Had to Die For years, OpenGL made real-time graphics accessible. It was a truly high-level ... a decent Driver-managed. in short Easy to get started. Then came the era of explicit APIs: Enter Vulkan. It arrived aaaand removed the magic. What Makes Vulkan Different? Vulkan is low-level, explicit, and built for modern multi-core CPUs. Unlike OpenGL, it doesn’t hide work in the driver. You are now responsible for: Memory Allocation Command Buffers & Queue Submissions Synchronization & Resource Lifetimes That means: Less hidden overhead. More control. More responsibility. ⚙️ Why That Matters in Production In massive open worlds or complex DCC tools, driver overhead is the enemy. Vulkan enables: ✔ Better Frame-time Stability: Fewer "stutters" from hidden driver tasks. ✔ Massive Draw Call Scaling: Perfect for dense, high-poly environments. ✔ Predictable Memory: No more "mystery" spikes from the API. AS a Comparison OpenGL: “Tell the driver what you want; let it figure out how.” Vulkan: “Tell the GPU exactly how to execute every byte.” 🎮 Why Tech Artists Should Care Even if you aren’t writing shaders or C++, understanding the API changes how you optimize. It makes GPU profiling clearer and helps you understand why modern engines (like Unreal 5 or Unity) have specific "rules" for draw calls and memory. Vulkan isn't about being difficult; it's about being explicit. Modern pipelines from USD to real-time rendering are moving in the same direction: Less magic. More clarity. Infinite scalability. Have you noticed performance differences or "stability wins" after switching to Vulkan in your projects? #Vulkan #OpenGL #Rendering #GameDev #TechnicalArtist #Realtime #GPU #EngineDev #GraphicsAPI #Optimization #OpenToWork

3

It took the latest Opus 4.6 model to sway my opinion, but I’m finally starting to see value in AI assisted coding. We built a Ninja-based build system for a hybrid C++ & Go project in 3 minutes, from scratch. No existing reference material in the repository to compare against. No CMake involved at all (yay!). I told it what I wanted, and that was mostly that. It was literally one minute for Claude Code to analyze the project & generate the files. Then one minute for me to realize it didn’t work because the commands were in Bash shell format instead of Powershell format. Then fix and test. I’ve never brought a build system online so fast before. It took this experience to get me to finally come around to the idea that a 1-person company can make it to a billion dollar valuation.

33

4 Comments

I have been a big fan of Plandeck for a while but their AI adverts I see now are so bad. The voice does not match the person speaking and the are using the same AI generated guy for different companies, all they have changed is the shirt he is wearing which gives the impression these testimonials are fake.

3

1 Comment

When I started in games, it was considered wrong to use C - real games were written in assembly - coded close to the metal. While it’s true that assembly is always faster at runtime, it takes longer to write and is harder to use for complex programs. Today nobody writes the bulk of their game like that. At the time it felt the resistance to change was as much about devs sticking with what they knew as it was about the performance loss. Claude tells me the full journey from assembly to C to C++ took roughly 15 years of gradual, resistance-filled adoption. At the UKIE conference last week one of my key takeaways was that while many PC and console devs are highly sceptical of AI, mobile devs have embraced it wholeheartedly. The recent GDC survey suggests that over half of AAA and indie developers now say AI is having a negative impact on the industry. While AI will have some bad aspects, the motivation feels more like a resistance to change and a lack of understanding. Meanwhile mobile has adopted it across its entire pipeline - asset generation, localisation, player personalisation, ad creative, QA and more. Iterating quickly and shipping quicker. It's a joy to use - an idea for a game can come from an artist who uses it to code or a coder generating art and while it may then be rewritten, as a way of exploring concepts and empowering creativity in all aspects of games, it’s a game changer if you’ll pardon the pun. ( even BD - it can make powerpoint slides like a champ.) AI won't replace great game developers. But developers who embrace it will replace those who don't. (If you’re lost on where to start, open a claude window, tell it what you do for a living in text or voice in as much detail as you can and describe how you like to work and ask it how it can help. Or if you're in Malta next Tuesday come and visit https://luma.com/0frezkj3) Image stolen from @joysilvart

30

12 Comments

I wanted to write some more thoughts after last week's post (that unexpectately went quite viral!) on the topic of devs not updating their JIRAs regularly. This seems like it's quite a universal experience no matter where you look. When there's so much friction against something so simple, one has to ask themselves, what are doing it for? My personal opinion on the topic is this: JIRA tasks (and other issues types) are intended to replace physical cards and post-its that live on a physical board. At certain scales you simply cannot have everything run on a board, hence the digitalization of our tools. A card (task, user story, epic whatever you want to name it) is supposed to be a conversation starter. Not all requirements need to be on the card. That's the whole point of using small cards, so you cannot write down all the specific requirements. I have a feeling that JIRA and its tasks, in a lot of places, only serve to provide PMs with data to see where there project is headed. (and devs understand this. The tool doesn't serve them so they don't interact with it). This is also quite natural I think. Projects are often complex and getting accurate progress information is hard to get by. You would need to rely on a very solid communication structure to surface that information. And that's incredibly tricky to set up and maintain. So JIRA often ends up trying to replicate this communicating structure. And that's where things go downhill in my opinion. Once your JIRA database becomes your main tool for work communication it stops being a work tracking tool. I believe that we can do better by utilizing JIRA in the same way that we would use a physical board. And our issues in JIRA the same way as we would use cards and post-its on this board. Keep things short, see each issue as the start of a potential conversation, make sure your teams keep talking, track the data and stay realistic about what JIRA is. It's supposed to be a representation of the work, not the work itself. And because the work is so complex, the tool has to be a simplified version. The map is not the terrain. Happy to hear your thoughts in the comments below.

502

56 Comments

Stream & Review CAD Models with the Unity Industry Viewer Template In this technical walkthrough, you’ll discover how to set up and work with the Unity Industry Viewer template — a plug-and-play starting point for creating collaborative, cloud-enabled 3D review experiences. The tutorial guides you from a blank project all the way to streaming your own 3D models for real-time team reviews. See what's possible: ✅Create a new project using the Industry Viewer Template from GitHub. ✅Connect your project to Unity Cloud to enable services like Asset Manager and multiplayer. ✅Prepare complex CAD and BIM models for 3D data streaming. ✅Load and explore your 3D models streamed directly from the cloud. ✅Use built-in collaboration tools like 3D annotations, section cuts, and XR support. The Industry Viewer Template offers a flexible workflow for teams that need to visualize, evaluate, and collaborate on complex 3D data in real time. It can be customized for AEC, manufacturing, automotive, and other industrial use cases. 🌐 www.progressive.com.my ✉️ pcssb@progressive.com.my 📞 03 7727 5139 #pcssb #pcssbmy #progressivecomputersystems #Unity #Unity3D #UnitySoftware #AEC #Architecture #Engineering #Construction #Manufacturing #CAD #BIM

3

In the UK during GDC week, with a nice clear diary. So we have, for a while, reached the episode of “LinkedIn is Facebook” where people who check it every day/multiple times a day, only post about how uncool it is to use it. It’s late-stage social platforming and it’s not just LinkedIn (in fact, in truth it’s not just social platforms, Steam-first marketing is harder to see results from). Fatigue and fragmentation push people toward trusted sources and direct relationships. Teams that invest in owned and earned media, communities, and partnerships weather platform volatility far better than those that rely on single or limited silver bullet channels. We have good case studies on how you can use smart cross-channel strategies to drive wishlists and sales. Get in touch if you want to hear more, this is the perfect week for a conversation.

6

2 Comments

🚀 Introducing ComfyEnv – An Environment Manager for ComfyUI ComfyUI is incredibly flexible, but installing different custom nodes or dependencies often leads to conflicts that can break existing setups. ComfyEnv helps you manage isolated environments, so each project or node set can live in its own stable sandbox. GitHub: https://lnkd.in/d5jgQ4Jd Try it, break it, and feel free to contribute or drop feedback!

20

1 Comment

Unreal Engine and I have had a very long, love-hate relationship for over two decades now. I am starting to enjoy working with it more and more since the 5.x release, but the elephant in the room is, and always has been that testing is a bit of a bolted-on, second-class citizen. I recall the Herculean effort at Rare to get comprehensive test support in (this was back in the 4.x days, but not a lot has changed). Yes, Gauntlet and some other heavier support frameworks are in place on one end of the spectrum, and simple automation exists on the other. I cannot place the blame at Epic's feet for the situation. There has been, across the industry, a reluctance to make any investment up-front to ensure games are "testable" and zero appetite for it once the tens of thousands of lines of "throw-away" prototype code suddenly becomes "production". An approach I have been taking is to start with tests first. Simple automation doesn't cut it once any GameMode or player controller code needs coverage. Gauntlet is a bit heavy. Fortunately, Epic does provide FunctionalTest controllers. The cleanest approach I have run across so far is to create a test plugin that only builds with the editor, is never packaged with the client, and can be run using automation from the command line along with OpenCppCoverage to generate a coverage report that any CI/CD system can run. Since it is a plugin, it can have it's own content/maps folder. I create an empty map that contains only a functional test controller. It is reasonably fast and has all of the engine and world setup required to properly test game code. I ALWAYS start with my CI/CD test script requiring 100% coverage or the build fails. Leaving that path is easy -- getting back onto it is 100x more difficult, so better to stick with it. I prefer this to more heavy-handed frameworks because it is very easy to setup and maintain early in the project, but robust enough to carry it through prototyping and even in to production. If any TDD purists out there have better suggestions, I am all ears. There is very little on the 'net covering this topic. I may update my "EmptyProject" template and share it publicly on GitHub at some point to include an example of this approach. It would be nice if the default project setup from Epic included something like this to facilitate better habits in the industry.

29

7 Comments

I'm heading to GDC next week, here is some of the exciting things Nequinox would like to share. At Nequinox, we build things thoughtfully, technically sound, scalable, and built to ship. Over the years our team has helped deliver multiple AA and AAA titles across console and PC, tackling complex engine, performance, and platform challenges along the way. Today that focus includes: ◾Co-dev strike teams supporting ambitious and fast-moving teams using our proprietary tech stack (NFUSION) ◾Our build & performance suite (COPHIRE), reducing integration friction and strengthening multi-platform stability ◾Original IPs ▪️Turbo-Tanks (8-player competitive racer, EA-ready this year) ▪️More Than A Tempo (A narrative RPG that trades combat for DJing set in the underground UK Grime scene, funded by the UK Game Fund) ▪️Trap Runners (An asymmetrical multiplayer where Trappers and Runners are against each other to complete an obstacle cause filled with deadly traps) We specialise in optimisation, console porting, performance turnaround, and multi-platform development in Unreal and custom engines, while helping teams build smarter, ship cleaner, and raise technical standards across the board. We’re investing heavily in next-generation development workflows - from automation and performance intelligence to practical AI-augmented development tools and production - helping shape what modern game development becomes, whatever form that ultimately takes. After nearly three decades in this industry, I’m still excited by where technology is taking us - and by the teams pushing to build the next generation of games. If you're at GDC and are: ◾A publisher looking for well-scoped original IP ◾A studio needing experienced technical reinforcement ◾A team ready to strengthen build and performance infrastructure Let’s connect! Always open to thoughtful conversations about making games smarter. See you in San Francisco. Our Website: https://lnkd.in/eAuvhWPD #GDC2026 #GameDevelopment #CoDev #Publishing #UnrealEngine #GameTech

55

Following the Loki sneak peeks, I’m incredibly excited to finally share something that’s been quietly in development for a long time. 🚀 OpenDCC is now public — the extensible, USD-native application framework behind ShapeFX Loki — and it’s now open source. This started with a simple idea: What if we could build a DCC that feels at home for artists — but gives studios full control over what powers it? I’ve personally been waiting for this moment for more than 7 years — and it’s hard to describe how satisfying it is to finally see it live. After years of prototyping, rewriting, and learning from real production needs, the framework is out in the open. It’s still evolving — but it’s real, usable, and ready to build on. And for those who’ve been following this journey — at SIGGRAPH, in the OpenUSD community, or across developer circles — we’re finally delivering what we’ve talked about for years. 🔧 What’s inside: • Extensible core built on OpenUSD & Hydra • Qt UI with dockable panels • Embedded Python (Pybind11 & Shiboken) • Structured package system with C++ & Python entry points • Command system that records actions as executable Python scripts • Standalone RenderView app (OIIO, OCIO, ZMQ) • VFX Common UI widgets ready to use • A growing library of example tools and plugins • …and many more goodies inside 🎁 Already includes all Loki v1.1 sneak peek features: • HydraOps (node-based USD scene composition) • Live Share (real-time multi-user USD editing) • Physics Layout (Bullet simulation) • Animation Curves (timeline & graph editor, baked to USD) 🛠️ OpenDCC is still a work in progress — but it’s live, open, and evolving. 🙏 Huge thanks to all the engineers and artists who contributed along the way. We did this, guys. 💡 And deep appreciation to the Pixar team for continually evolving OpenUSD — this wouldn’t be possible without that foundation. Explore what we’ve built on top of OpenDCC: https://shapefx.app Browse the code on GitHub: https://lnkd.in/e_rsmMnh #OpenDCC #OpenUSD #DCC #VFXPlatform #VFX #Hydra #USD #ShapeFX #Loki #OpenSource #Animation

94

7 Comments

Unreal Engine 5.6 Preview is out! Looking forward to trying it out! A summary of the new features: 🖼️ Rendering: Lumen Hardware Ray Tracing Optimizations, RHI - Renderer Parallelization, and Virtual Shadow Maps Optimizations. 🧍‍♀️ Character & Animation: Sequencer Organization, Seamless Gameplay Transition for Cinematics, and MetaHuman Creator in Unreal Engine (Beta). 🗺️  World Building: Includes Fast Geometry Streaming Plugin (Experimental), Landscape - Batched Merge, and Procedural Content Generation. ⚙️ Developer Iteration: Features like Incremental Cooking (Experimental), UBA Cache Server (Experimental), and New Project Launcher UI (Beta). 🎬 Motion Design: Cloners & Effectors, PSD Layer Import, and Motion Design Scene State. 🎥 Virtual Production: Nvidia DLSS Support in nDisplay, Take Recorder Subsystem, and Production Wizard. I am personally very exited to try the New Project Launcher UI as soon as I finish to download the preview. What about you? #UnrealEngine #UE #GameDev #Rendering #Development

14