Adrian Sanchez 🎮

Don't let em tell you you can't have crispy clean shadows on the web 😉

Social media is where the sensible takes and opinions go to die.

When you get to this point you've effectively transcended into the realm of a researcher, because you have to actively try to figure things out yourself with very little guidance, and in the process you may discover something no one else has.

If you look at AI (ML) in games, it shows that it can work exceptionally well for bespoke things. That's probably where continued research in game-based ML should go. You would probably get a lot of mileage out of using many small models to do different things internally than a…

The fact that people see AI models as their own intelligences instead of as an extension to human intelligence is still odd to me.

Does anyone who makes indie games actually make good returns on their games? (Not counting the needles in the haystack like Minecraft, Balatro, etc.) It seems like people making tools usually have much better returns on average (e.g. many in the SaaS world). Maybe indie game…

The evolution of humanity and technology is likely stifled by compartmentalizing things too much. The fact that gaming and AI diverged from each other as much as they have was probably an accidental mistake. I can never quite find the right words to describe my vision for this,…

Are there any good resources on building/updating clipmap-based dynamic voxelized SDFs (via tri-meshes and/or edit lists maybe)? Thats sounds super fucken specific so probably not.

"Slop" is a less rosy way of saying you think something is subjectively bad. There is no objective "slop". Any piece of AI slop is a distillation of all the human-made creative works it was trained on. It's a distillation of all human culture, history, and art, and it can only…

This is the way

Clear evidence AI can help you A/B test visions for games. The moment I saw this, it just clicked. There's easily a few million dollars for anyone who can make a compelling, open world game in this aesthetic.

If you're a software engineer or have worked on completing solutions to complex problems at all, there's this funny thing that tends to happen where you feel like you're 80% to 90% done with the solution, and can almost see the finish line, but there's still some glaring problems…

An incomplete-content-addressable memory machine is probably the best description for neural networks I've heard to date.

Discourse in AI has become more interesting than discourse in games. Games discourse is now just bs politics, how everyone hates executives, layoffs, and barely topical speeches about the 5678th FPS or RPG that looks like any other cookie cutter game. Barely any talk about…

Once again, it's full circle. Video game technology inevitably leads to AI technology and vice versa. The underlying, unspoken motivations behind both are inextricably linked. There are no coincidences here, and this trend will continue into the future.

Caved and spent a few days adapting the XeGTAO implementation into Sundown. Still not perfect but good enough. Bent normal support is also only half there and not enabled yet. (Left: GTAO, Right: No AO)

I think one of the biggest flaws with our understanding of the universe is this: we think physical laws and the laws that govern the universe are fixed and intangible, forever. I think what's really happening is that much like biology and brains, the universe is just a giant vat…

It's almost GI time