Dan Miller
@dan0mighte.bsky.social
91 followers 170 following 170 posts
game dev. Software engineer. Systems obsessed. he/him check out my work and blog: http://linktr.ee/fraculation
Posts Media Videos Starter Packs
Reposted by Dan Miller
theduckcow.com
That's one small step for [mesh], one giant leap for [road].

Procedural intersections are officially in progress, major kudos to Picorims for the work thus far on this feature! #godot #procgen #gamedev

Details, if you are curious to follow along:
github.com/TheDuckCow/g...
Image of a road in a 3d godot editor looping around and connecting into a white triangle mesh without any materials or smoothing
dan0mighte.bsky.social
Holy shit. That looks very sigmoidy to me!
dan0mighte.bsky.social
I wonder how sudoku would work. Wikipedia states about 5 billion unique sudokus under certain symmetries. Maybe a few days of compute could graph that. But a node for each time a cell is filled in... that would explode
dan0mighte.bsky.social
Right, this thing really gives you a sense for the exponentials! I thought sokoban would be great for this at first, but practically every addition to a puzzle doubles the state graph. I'm hoping interesting puzzles still exist even with low node counts
dan0mighte.bsky.social
I wonder what the graph of chemical activation VS light exposure looks like for silver halide crystals 👀
dan0mighte.bsky.social
OH I see! You get SUPERHDR. It sounds to me like you are modeling physical reality more closely than traditional path tracing... It'd be poetic if the answer ends up matching the real world: more ambient light, scattering, etc. Or somehow replicating how actual film does tone mapping
dan0mighte.bsky.social
These are beautiful. What I assume is low ray sample count reminds me of a film grain style effect
dan0mighte.bsky.social
Those dang bot boats.....
dan0mighte.bsky.social
My post-hoc theory: Branches represent choices between different boxes. The initial state is somewhere in the center. The player can pick any of the 4 boxes to move to the goal at first (4 trunks coming off the center). 3 trunks come off each of those, representing the remaining 3 boxes, and so on
dan0mighte.bsky.social
@limeblossom.bsky.social gave me the idea to treat each box as unique objects rather than as interchangeable with each other. This creates more states so its a bit harder to analyze. But it also creates a ton of symmetries which I think make it quite pretty!
dan0mighte.bsky.social
Got a cool video capture of the graph solver running now that its optimized into the range of realtime
dan0mighte.bsky.social
It is interesting how alt-text labels for images is almost exactly what a generative AI needs to train. Not advocating to omit, but, I always have this gremlin thought in my mind as I write them out
coil.bsky.social
Irregular reminder that alt text on images is a valuable accessibility tool, and also lets your images show in search results or respect mute lists. For instance, I found the below image by searching "capybara alt text".
A brief guide on writing good alt text for images, which suggests including 
1. Who or what
2. Their expression or emotion
3. A description of the image
4. Notable colors
5. Interesting features

The included example image and description: "A capybara looking relaxed in a hot spa. Yellow yuzu fruits are floating in the water, and one is balanced on the top of the capybara's head."

Dan's addition: saving the "interesting thing" for last is a bit like writing a joke - end with the memorable punchline!
dan0mighte.bsky.social
conveniently, due to the nature of the simulation bi-directional edges attract stronger than single-direction. This chain is composed of small groups which you can navigate between easily, but it appears one can only progress in one direction between the groups. (without hitting a dead end)
dan0mighte.bsky.social
I made it 3D with a custom solver, and its starting to look promising. I'm curious about whats going on in this little chain separate from the rest of the mass. There must be something interesting here!
dan0mighte.bsky.social
This is precisely the motivation for it, nice catch!! I hope to reach as incredible visuals as they were able to achieve. I've been thinking about that video for what feels like months
dan0mighte.bsky.social
I'm hoping this could be useful to analyze if a puzzle is "good" or "hard". I'm not seeing any useful patterns yet. I'm hoping to find something like a choke point or some other graph property which can map to something interesting about the puzzle
dan0mighte.bsky.social
This level is only a little bigger but its state graph is enormously larger (9000 nodes vs 30). The player has many more unique ways to solve because there are fewer dead-zones for the boxes
dan0mighte.bsky.social
Playing around with analyzing how many ways there are to solve a simple Sokoban. Orange = no boxes on destination, Blue = all boxes on targets.
A 3D graph showing a mesh containing nodes of color ranging from orange to blue, with something that looks like a gradient across it A small sokoban level containing 3 blocks and 3 destinations
dan0mighte.bsky.social
I know some technology that is great at creating synthetic contributions....
Reposted by Dan Miller
ian-sundstrom.bsky.social
🔴, a #pico8 bounce shooter, is out now!
Play it in browser on itch: iansundstrom.itch.io/redcircle
Development fun facts itt 🧵
Reposted by Dan Miller
pailknight.bsky.social
Day 1! Coin inspired dress ended up more like scales, but dragons like money too, right?
Reposted by Dan Miller
ian-sundstrom.bsky.social
Going to release 🔴, a #pico-8 bounce shooter with mostly procedural animation and gfx, sometime this week!
Reposted by Dan Miller
southworth.bsky.social
I love environmental storytelling in dystopian sci-fi games.
nygroove.nyc
Yeah, everyone hates the AI subway ads as much as you do nygroove.nyc/ai-subway-ads/