elvis retweeted
Criticizing developers for using LLMs without understanding how they're actually used doesn't make you an advocate for better development.
It makes you part of the problem!
Big Walk used AI in 'the final stretch' of development, kicking off intense awards debate:
aftermath.site/big-walk-ai-i…
Readers added context they thought people might want to know
House House used LLMs sparingly near the end of Big Walk's development, mainly to fix bugs in third-party code. They estimate well under 0.1% of the game's code was affected, and they didn't use generative AI elsewhere.
bigwalk.game/faq/#ai
pcgamer.com/games/puzzle/b…
elvis retweeted
It’s absolutely essential that we maintain some degree of sense and reason around this. Talking to an LLM about a bug in an engine is completely different to AI generating all your art assets or telling claude to clone someone else’s game with slight tweaks.
elvis retweeted
This interview with GabeN is quite interesting:
“Perhaps unsurprisingly, Newell believes that LLM tools are the next great technological leap, and are going to prove so transformative that this is where people should focus their learning.“
pcgamer.com/software/ai/gabe…
elvis retweeted
Our game idea got stolen months ago. I know who did it and I know why.
Turns out you can steal an idea but you can't steal the skill to make it good.
The only thing that worked was their paid marketing. And the more ads they ran, the more our wishlists went up.
So thanks for the free advertising.
elvis retweeted
Average Unity asset flip experience
-> Planes flying sideways
-> Games so bad Valve had to call them “fake games”
-> A Unity splash screen so stigmatized it made the engine’s adoption stall harder than an Unreal Engine stutter
elvis retweeted
Follow-up of my last (hotly debated) poll.
Should fully and partially AI-coded games (but everything else is hand crafted) be required to have an AI disclosure on Steam?
50%I'm a dev. NO.
36%I'm a dev. YES.
6%I just play games. NO.
8%I just play games. YES.
2,832 votes • Final results elvis retweeted
Unity Asset flips used to be a thing.
Once the quick and easy money crowd gets burned they leave after releasing their first game.
There’s been a lot of discussion about the viability of Steam Next Fest recently, and first of all, I strongly recommend reading all the historical deep dives on NF data by GameDiscoverCo and How To Market A Game.
Yes…NF has become ridiculously crowded, but what if the median WL gain hasn’t actually changed that much in the last 6 years? It’s easy to say that AI slop is killing your visibility but is that actually the case?
For starters, the number of NF games that get almost no attention has always been very high. Even long before this AI slop era. Steam randomizes the games heavily for the first few days, but even then NF typically settles into a rhythm after the first 24 hours where all the discourse tends to be about the same 25ish games. As usual, the cream rises to the top, but having some marketing momentum behind you certainly helps!
So what has changed?
I think the gamers have.
NF 5 or 6 years ago was a much more niche event. There were fewer games, but there were also fewer players. I think players back then were the sort who enjoyed playing demos, offering feedback, submitting bug reports, and actively digging around for hidden gems.
Now, due to a number of factors such as increased discussion around NF, more streamer coverage, and players being more open to trying smaller games (friendslop era effect?), I suspect NF has seen a pretty considerable influx of new players.
And where do these players end up? Well, naturally, playing the top demos!
We love to talk about the highs and lows on Twitter. The games that explode and get 100k WLs, and the games that didn’t even break 100. But those are the extremes. In between those are thousands of games, and the data suggests that a huge proportion of them gain only a few hundred WLs at most.
That’s the reality of NF, and it has been for a long time.
We want to blame AI, and sure, NF does have a lot of AI slop, but are gamers really that dumb? Is anyone actually playing this stuff? I suspect the vast majority of it sinks to the bottom and rots almost immediately, just like low-effort games did long before AI existed.
I’m not saying that discoverability isn’t getting harder, or that AI slop isn’t a problem. It absolutely is. But I do think we’re far too quick to brand every disappointing result on the number of games joining NF.
If the number of games has exploded, but the number of players has also increased… and the median WL gains haven’t really collapsed, then perhaps NF isn’t becoming as bad as it might seem?
There’s a lot of data out there and a lot of nuance to the discussion, so I think it’s just worth approaching carefully.
I know it seems discouraging, and I know it might feel like you should give up, but don’t. NF has always been a gamble, and it’s a gamble I think is still worth taking.
The algorithm of steam already sorted most of it on next fest. Eventually people will see these slop/clones will not be profitable to justify making and $100 to publish them.
How much slop can you publish until you have a hit?Probably not enough.
elvis retweeted
the starting template for @kaigen3d is 400 lines of C code and gives you:
- a physics based character controller
- procedural skybox supporting day night cycles
- customizable editor
- runs natively on PC, mobile and web
you can point your AI agent and go nuts
elvis retweeted
9 months ago I shared a video where my engine beat @unity ECS on the web by 17x
I got a lot of criticism bc it turned out Unity didn't support Burst on the web at the time
they do now. so rematch:
ECS: ~2ms vs ~9ms
CPU: ~3ms vs ~28ms
@kaigen3d is still 4.2x to 8.7x faster 🧵
just finished my ECS system in C and it turns out it's ~17x faster than @unity DOTS on the web
that's 100k boids at ~90 FPS instead of 17 FPS
sharing the code and link for this demo soon, details below 🧵
elvis retweeted
we had a complex shader system for PS2. writing a shader meant an artist worked closely with an engineer to bridge the gap.
first thing i did for xbox 360 was write a material editor that generated shaders.
no longer needed expensive rendering engineers to make shaders and materials.
they weren't happy with me at all
elvis retweeted
NoGraphicsAPI is released (MIT license)!
It's a low-level Vulkan 1.4 wrapper using latest extensions (BDA, descriptor heap, device address commands, unified image layouts) to get as close as possibly to my No Graphics API blog post and SIGGRAPH 26 talk.
github.com/sebbbi/NoGraphics…




























