Phil Brown

@rompa69 Lowest common denominator between my iPhone and 45” desktop monitor 😝

What's with you web front end developers making all pages a sliver of the width of even the smallest widescreen monitor. You can't support variable width for both phones and monitors? Who are we kidding, most pages don't layout properly on phones either.

@rompa69 I relate to this my old friend 😂 old tech being rediscovered by new generation

Interesting that meshlet rendering considered new when it has been around since PS2 days? VU code would often process data in "meshlet" packet sizes to generate "optimal" GIF tags (rasterizer state) and also do cone based culling etc. Everything old is new again. I'm too old.

@SebAaltonen Aren’t they stack / bucket allocated by default now anyway? 🤷‍♂️

LLMs generating C++ absolutely love temporary short lived std::vectors and std::unordered_maps inside functions. So many wasted memory allocations.

Ok block man has a ray gun! Threw skinning and some basic IK so hands grab the gun. It's quite a mess still, but might have potential in an "abstract" kind of way 💭

Toying with the look in my GI ray tracer, my engineering brain wants to streamline everything and even make characters out of brushes 🤔 I wonder if this can work 🔥

@GlvGalv Very cool!

I wishlist you a happy new year 2026 : store.steampowered.com/app/4131730/VH…

Adding dynamic lights to my GI ray tracer, very cool to see the indirect reacting instantly. I also learned that ray traced shadows solves SO many pain-in-the-ass things that shadow maps have to deal with 🧠

@SquareGlade lol do you know what happened?!

Thank you so much for this Christmas present, 20k wishlist additions on a single day is absolutely mind-blowing! 🤯

Been fiddling with a global illumination (GI) renderer over the break. It's running at ~200 fps with 32 rays per pixel via OpenGL on my macbook, computers are fast! 😮

@ItsPolyCube Woa this is tight nice work!

I'm back from the break! Been really burnt out recently, so I didn't post anything for over a month. Now I'm better than ever and ready to continue with this game! #gamedev #indiegame

Slowly getting somewhere with Reflex MM, UI design it tricky. Going for minimalist yet clean look.

@SuperMegaTeam Looks great! The writing (glpyhs?) on the roads are great too, nice job!

We got feedback about enemies not properly being seen in the dark , so we gave them glowy eyes now. #screenshotsaturday #indiedev #roguelike

@rickyretouch So pretty!! What did You render this in? 🤔 👀 💭

10 years later this game still runs and feels great, memories <3 @ReflexArena #indiedev #arenafps 🚀

Just messing around with some physics 🙃

@nczempin @epyoncf @SebAaltonen Personally I’m a simple man and use Cursor purely for manual typing and ai auto complete (which is amazing)

@tallPhilb @epyoncf @SebAaltonen I second the Claude Code part; haven't really got warm with Cursor just yet. I am aware that you can essentially combine them. Not sure if you get to use a Claude Max subscription on anything other than the OG Claude Code, though.

People say that one should not build their own engine. It's not possible anymore as engines have grown to be so big. But all of these big engines have legacy foundations which can't be replaced as there's so much code on top of them. You can definitely build a better foundation today for modern hardware. And you don't need a big team for that. But what about leaf features A,B,C...Z? There's so many of them. A great core isn't enough. A useful engine needs all the leaf features. Same as OS vs kernel. This is where AI steps in. I have been doing AI assisted programming lately and it's actually very good at writing leaf code. You don't need a massive team. A small team of seasoned experts with AI agents is very productive. I know that most of my followers despise AI written code, but it's improved drastically in past year (again) and it continues to improve rapidly. Build the core tech yourself, design good architecture and good APIs and let the AI build the leaf features on top of your APIs. That's a solid way to build a hobby project too nowadays. You write the most important code (= innovation) and let AI write the less important code. A future game engine could be a thin core (kernel) with good APIs. Leaf features are modular and built by AI. Customers can build their own features on demand using AI.

@epyoncf @SebAaltonen Claude code (command line) and cursor are both leagues above for coding than copilot and ChatGPT

@SebAaltonen What are people using for that successful AI assisted coding? I do engine coding in both C++ and FreePascal + gameplay coding in Lua, and most of the time AI wastes more of my time than it saves. Been using Copilot and paid GPT. Anything out that is a significant improvement?

@SebAaltonen I’m way faster in my own engines than thethe alternatives. And this is before AI which have tipped the scales even further. 💯 agree

Building a level is so satisfying💚 Wishlist -> gooey.gg 🎥 youtu.be/u-fUuPL3e_g #GooeyGame #indiedev #gamedev #platformer

@SebAaltonen I went M2->M4 and it’s an amazing machine. I haven’t noticed the display latency, and I’m a latency nut, can you give more specifics? I’m curious 🤨

Going to skip M5 Max. It's more powerful than my M3 Max, but this time I am doing the upgrade for me. Last time M1->M3 Max halved the compile times = business reason = save money. M3 Max is almost perfect. The only problem is display latency. M5 Max doesn't fix it. M6 does.