Rosy

[cloud rivet]

I've been using xbox controllers since forever, the last non-xbox controller I had was the ps2 controller

in 2003

my console ownership history is sega game gear -> virtual boy (lol) -> play station -> play station 2 -> xbox 360 at which point I had became a dad and mostly stopped playing games

oh I had a gba for bit and a nintendo 3ds but I never used them much

the nintendo 3ds was horrible

[tight torrent]

tbf its a pretty good controller

and its the xbox series one, not the xbox one one

[tight torrent]

no its not 😡

[cloud rivet]

it hurt mey eyes, physically

[tight torrent]

I've owned mine for 14 years

[tight torrent]

Oh yeah the 3d is dogshit

[cloud rivet]

the virtual boy also hurt my eyes but I loved it

[solid grove]

i didn't have handhelds growing up so I don't have the nostalgia for them. I played a real DS and 3DS recently and it was not a good experience

[tight torrent]

other than that, fucking phenomenal console

[cloud rivet]

it was my not playing game phase and every game I played made me want to turn it off because of eye pain

so I never tried it as just a 2D game thing

[tight torrent]

yeah the ds/3ds is a bit small, thats why i have the circle pad pro thing

when i was younger it was great, i had much smaller hands

there were many a mario kart grand prix played on this 3ds

but now i cant play it without the extra circle pad

[cloud rivet]

oh I had a sony psp or whatever too I forgot

[solid grove]

IME the screen is also low res and kinda crappy, the speakers are super crappy, and it's uncomfortable to hold

[cloud rivet]

I never played that much either

[tight torrent]

all fair, the games really carried the console imo

[cloud rivet]

I kept wanting to play games but just spent all that time being a dad, which, I think was better time spent

[solid grove]

play games with ur kid and get both at once

[cloud rivet]

yeah she was too small then, we played games together when she got older, we played PC games though

she mostly played with her friends though which is good

we played overwatch mostly

[broken fog]

you take that back immediately

dualsense

the ps5 controller

[broken fog]

psp is far superior ergonomically, ds/3ds is for double screen memes

but it's a pretty great handheld with an absolutely fantastic library

screens are low res and crappy but meh who cares

someone shared this thing in offtopic a while back tho and it looks neat https://www.ayntec.com/products/ayn-thor

[astral hinge]

I played call of duty on my DS

even played multiplayer on it

[cloud rivet]

the DS came out after the 3DS?

[astral hinge]

no

[cloud rivet]

they took the 3D out?

oh

ok

[cloud rivet]

looks pretty

[broken fog]

they did make the 2ds

which was after the 3ds, and took out the 3d

essentially a "budget" 3ds

[tight torrent]

the 3ds has monster hunter 3 and 4 ultimate, mario and luigi dream team and superstar saga remaster, mario 3d land, pokemon gen 6 and 7 and gen 3 remakes, zelda OoT and MM remakes, zelda ALBW, AC:NL, MK7, smash 4, and support for every DS and DSi game

[tight torrent]

basically yes, the first version of the 2ds also removed the clamshell, it was just one big flat brick

[broken fog]

don't forget like three fire emblem games and a ton of good jrpgs

[tight torrent]

exsactly

[cloud rivet]

3ds kinda killed me being a fan of nintendo too lol, I'm sorry. I tried to like it, but it was not great for me

[broken fog]

should be able to emulate 3ds and psp easily, i'm kinda tempted to get one actually

[solid grove]

nowadays i can only play games on a big screen. steam deck isn't for me either

[tight torrent]

the later refresh of the 3ds (the "new" 3ds/2ds line) just took the new 3ds and removed the 3d effect

[cloud rivet]

will you make your game work on it?

[solid grove]

considering that I do not own a steam deck, no

[cloud rivet]

I think you can get a free one

[solid grove]

my engine also completely breaks proton so uh

[cloud rivet]

oh

[tight torrent]

how lol

[solid grove]

something I'm doing causes it to draw confetti for a bit then crash

but native linux builds work fine

[tight torrent]

💀

[cloud rivet]

I'm just making my game for me, so it's fine if it just runs on my pc

[broken fog]

then it should run on deck? no reason to use proton if the game supports linux natively right?

[solid grove]

well when I had access to a steam deck, steam just refused to run the process. but I've tested native linux builds on other linux systems and they worked

[broken fog]

idk don't have a deck either

i have the opposite problem with it though, too big

[tight torrent]

i have a deck, its very nice

i still game a lot not at my house, by nature of being a college student

so deck + switch + laptop does whatever i need it to

[pseudo dock]

The thing that set the DS apart for me were not the traditional games (so, like, the Mario 64 port for the DS wasn't very fun for me to play), but rather all of the unique games for it. I think the Phoenix Wright and Professor Layton games are probably the most famous, but there were a lot of really interesting games like that that were different from what was being released on standard consoles.

[cloud rivet]

hadn't heard of those

[pseudo dock]

Maybe you'd recognize a Phoenix Wright meme and just don't recognize the name

[cloud rivet]

oh yes, I recognize the meme

[tight torrent]

theres alot of really unique games for the ds and 3ds as well, like the zelda ds games (phantom hourglass, spirit tracks) have very unique control schemes iirc

[tight torrent]

THE ALL TIME PEAK OF PHOENIX WRIGHT ACE ATTORNEY TRILLOGY ON STEAM IS 4,000 PLAYERS

(wait no it has like a million sales shit)

[pseudo dock]

So there's an interesting example of the magic of the DS that I don't quite understand myself: I love love love the first three games (I like all of them, but the first three especially) and bought them on Steam, but I've never played the Steam versions. Something about the form factor of the DS and the kinds of games made me play it a lot, but I never really think about playing it on my computer (or even my Steam Deck).

[tight torrent]

thats completely fair, i dont think i could play kingdom hearts dream drop or mon hun 3 or 4 on emulator

ive also played alot of pokemon on it lol

experiencing all the gen 7 cutscenes without skipping them with emulator hacks is a special kind of hell that can only be experienced on console :)

anyways, im gonna go play the new pokemon game that released like

20 mins ago

[solid grove]

and in another 20 mins it'll render its first frame!

[cloud rivet]

:O

I have never played a pokemon game

[tight torrent]

i uhhh

have played probably too many

[tight torrent]

its actually better than SV

performance wise

[cloud rivet]

ok I got my orientation code merged, now I want to get just like one physics thing in,

linear dynamics ya this looks good

[solid grove]

some of my friends made an f-zero type game in uni and from what I remember they raycasted "down" to the track to get the closest triangle of the collision mesh and then used its normal vector as the current gravity vector

[cloud rivet]

oh interesting

I actually don't know how I want all that to work

I think maybe I just go with normal gravity and just when you go really fast around a turn you can stick to the wall a bit but if you slow down you fall I'm not sure, I don't really know, whatever will turn out to be fun I guess

I just have to get something to move at all right now

[solid grove]

lol this is what I had back in june https://cdn.discordapp.com/attachments/1173035001554677863/1381849899900866623/ai-drivers.mp4?ex=68f1c381&is=68f07201&hm=75f1e33af1dc6d41b4ddd9f05a4784df6a35201567b2fbb1d86c8b02515e193e&

[cloud rivet]

amazing

it's pretty late, I don't know if I'll get anything in today, but the moment I have movement I basically have gameplay

[solid grove]

walking simulator devs be like

[cloud rivet]

Tony Anderson's been my coding music for a bit https://www.youtube.com/watch?v=zDqvlvXn4TY&list=PLj5k56xuVS2FYCbh9czr0tcSJc2HDEtFY&index=1

[cloud rivet]

signs of life I guess, idk, lot of work to do

[cloud rivet]

I haven't gotten around to updating the TLAS yet

I'm going to work on that tomorrow

but a little bit of progress today

but the shadow gets left behind 😅

[solid grove]

pro tip: don’t try to use an empty BLAS in your TLAS, it will crash the GPU. took me forever to determine this

[cloud rivet]

oh good to know

after I fix the shadow/tlas issue I'll work on the physics more

I need to support rotational dynamics and add friction and force

and then I will make the track an actual thing, right now it just is hard coded in the mesh shader but I will need to have the spline define the track in the application so I can use it as a ground and not just hard code the xz plane as the ground

then the vehicle can fall off the track

then I need to add the third person camera view for the vehicle

and then it's just working on the track generation and vehicle behavior

I'll just hard code track tangents and points, I don't want to spend any time on an editor for the first couple of tracks unless I think that will save me time, or maybe I can use an external tool to create splines and record the points and tangents

that would be amazing

[solid grove]

I make my track collision meshes by hand in blender

the scenery is just for display

[cloud rivet]

I was thinking of just using spline math, but I could just make the tracks in Maya

or Houdini

I think Houdini would be better

that'll be fun

and then I export as obj

yeah the scenery will just be for display also, the scenery will be RT pipeline

it'll be completely static

anything that moves will be graphics pipeline and RT shadows, I might create optimized shadow meshes

I don't know how I'll do IBL or GI

I'm not worry about any of that yet

I'll need occlusion information from the scenery though

I think I can generate it

[cloud rivet]

going to be using this to control the vehicle

:D

[hushed creek]

Nice

On PC I'm still using my trusty logitech f310

It's like the one they used on the titan submersible but wired because I'm not crazy

Actually I saw the f310 in a lot of random places in some video they were controlling some rover with it at nasa or sth

It's cause it's a cheap controller that is wired and also good enough

[solid grove]

I hav the Xbox One version of that controller

[hushed creek]

Well what might this be if not a gameplay idea

Hmm

[pseudo dock]

Oh hey, a chance to plug a possibly-not-well-known game that I liked despite its flaws that involves interesting gameplay with shadows:
https://en.wikipedia.org/wiki/Lost_in_Shadow

[hushed creek]

I was unaware of it, nice

[cloud rivet]

I'm going to find an anti grav racing game on steam and see how the controls are handled

there's gotta be a billion of these

https://store.steampowered.com/app/1954550/?snr=1_5_9__205

Steam

Project Hover

Description

Welcome to the world of high-stakes arcade racing with a twist! In our game, winning isn't just about crossing the finish line, it's about claiming your rival's antigravity-powered racing machines. Dive into a rich world of factions, each with its own style and controls, and push your skills to the limit with our intricate anti-g…

Price

$10.99

▶ Play video

https://store.steampowered.com/app/2486820/?snr=1_5_9__205

Steam

Sonic Racing: CrossWorlds

Race across land, sea, air, and space in Sonic Racing: CrossWorlds! Warp through Travel Rings into new dimensions where something new awaits around every twist and turn.

Speed to victory solo or as a team in a variety of offline and online modes and compete against players from around the world. Build the ultimate vehicle to match your racin…

Price

$69.99

Recommendations

6168

▶ Play video

https://store.steampowered.com/app/2078450/?snr=1_5_9__205

Steam

Warhammer 40,000: Speed Freeks

…

Price

$19.99

Recommendations

253

▶ Play video

there's not a lot honestly

a lot of traditional racing games or mario kart clones

I'm going to try the hover one

[cloud rivet]

That's like, an unfinished game with broken controls and 2 maps lol

[pseudo dock]

So you'd recommend it, then?

[cloud rivet]

not for $10 lol

no not really

no updates in forever

just looks like a hobby project

sonic racing has a demo I will try that

the game costs $70

so

not sure I will buy that one

[pseudo dock]

Did you learn anything? Maybe at least anything not to do?🫠

[cloud rivet]

it's not really want I want to make

it's too much like a car racing game

I think the trailer for it is pretty cool

[pseudo dock]

I don't know what qualifies as an anti grav racing game, but AniCator made a racing game for a game jam that you could check out for free 🙂 #1126099727621046332 message

[cloud rivet]

yeah I played it

idk I'm just making a small game for me to work on my engine and learn things

[broken fog]

maybe lean on the complete lack of friction with the ground

and make the whole thing about controlling inertia

hover drifting

[cloud rivet]

as someone not good at these games or any video games I'm just doing this for me and whatever works and is fun for me to do

[broken fog]

yea, that's fair

just suggesting if you don't want it to feel too much like a car game you can do a lot worse than leaning on the unique physics of it

[cloud rivet]

yes that makes sense

[cloud rivet]

I'm going to fix shadows today, then add the third party camera, and then I'll start working on an actual track so I can fall off of it if I \go over the edge

I'm not going to build anything in an app, because I just want to use a mesh shader for it and have it be defined by points and tangents

[cloud rivet]

all due respect to john khronos, but the RT API is very gross

I think I won't have to touch it again until I add skinned meshes though

that's definitely an API you want to wrap with helper functions while wearing a hazmat suit and full PPE

still better than using a CSM though

[solid grove]

why is the corner of the shadow beveled?

[cloud rivet]

the light direction

[cloud rivet]

I switched from Windows Treminal to WezTerm it's pretty cool, I couldn't take it anymore how Windows Terminal doesn't open any new tab from the previous directory

nvim looks so much better in Wezterm, and it's configurable with Lua too

and I installed it on my work mac too

https://wezterm.org/index.html

[cloud rivet]

putting Claude to work finding me example games, because I failed in my own search

[hushed creek]

the "thought process" part sounds so robotic XD

[cloud rivet]

I won't actually read the 10 page report it generated, I just give that back to the AI to break into a list

it took like 10 minutes and search 400 websites

I love this feature

can I paste markdown in DIscord

Aero GPX

97% Positive | $19.99
The closest spiritual successor to F-Zero with instant-response controls and speeds reaching 1,500+ km/h.

BallisticNG

95% Positive | $14.99
The ultimate WipEout tribute with 50+ tracks and 200+ community creations via Steam Workshop.

ArcRacer

100% Positive | $8.00
A budget-friendly pixel-art racer that runs on integrated graphics with tight F-Zero-style gameplay.

Redout: Enhanced Edition

85% Positive | $19.99
The fastest racing game ever made with pure physics-based racing and no combat.

Redout 2

81% Positive | $29.99
Even faster and more challenging than the original with speeds reaching 2,000 km/h.

GRIP: Combat Racing

87% Positive | $29.99
Race on walls, ceilings, and any surface at 767mph with heavy weapons and VR support.

Radial-G: Racing Revolved

77% Positive | $14.99
Authentic 360-degree tubular track racing built from the ground up for VR.

Pacer

72% Positive | $39.99
A competent WipEout homage from ex-WipEout 3 developers with pre-race weapon loadouts.

Antigraviator

64% Positive | $19.99
Trap-based combat racing with no speed limits but limited content.

Gravity Chase

Mixed Reviews | $14.99
4-player local split-screen tubular racing similar to Killer Loop's design.

oh yeah that works

these are great

the redout 2 link is broken https://store.steampowered.com/app/1799930/Redout_2/

lol gravity chase links to a cake shop simulator

actual link https://store.steampowered.com/app/1171520/Gravity_Chase/

ok redout and GRIP seem to be only real games, the rest are small projects with very few players

I'm going to try those three games

redout, redout 2 and GRIP: combat racing

these are exactly the kind of game I want to make

[cloud rivet]

wow Redout is amazing

[cloud rivet]

couldn't upload my video to discord of a timed track race, so put it up on YT instead https://www.youtube.com/watch?v=1-cc9XEWNRQ

I got a lot of work to do lol

I need to keep playing these games to get better at the controls

[tight torrent]

Yeah their both pretty cool

They were both free on epic a little while ago

[cloud rivet]

redout 2 is also great, really amazing looking maps, but seems more difficult to play

going to try the last one, GRIP and then I'm going to do some work

[tight torrent]

I didnt play 2 a whole lot, didnt play either a whole lot tbh

[cloud rivet]

yeah these don't seem like popular games

the indie games all seem DOA ghost towns with no traction

[tight torrent]

It kinda felt like a spiritual successor to fzero to me

[cloud rivet]

yeah

redout 2 kind of stuttered for me, I may have to close other apps or change settings or something

[tight torrent]

Yeah I dont think they are like at all

[cloud rivet]

both redout games are beautiful

I am just making this for me and to learn so I don't care, I like it

[tight torrent]

Oh absolutely

I have more fun when im playing racing games with friends or whatnot, and usually Noone wants to hear "hey wanna play a niche indie game"

[cloud rivet]

man GRIP is also really cool, it has weapons and the gravity mechanic is a little bit different

Redout 2 difficulty is probably too high for me right now, I love the maps. I think I will stick with the original redout to learn the controls

ok time to work

will add the third person camera

I'm glad I found these 3 games

it feels like a good demo game to try to make, the goal is to just follow a track, no complex game mechanics. the game challenge is in the physics and controls, which will be a real challenge to do well, but just like a basic game though this isn't overly ambitions. I can start with a small track and just get that working with basic graphics

[cloud rivet]

now I also have something to do when I need a break from programming that isn't trolling this server

it's nice you can just play a timed race for a few minutes, it's not like a multi hour investment

[broken fog]

those look really fun tbh

[cloud rivet]

yes I think so

also just so pretty

I think maybe it's easier to make these types of games look nice since the world is on rails

you can bake a lot, and you can do a lot of expensive things because the player 's movement is restricted. Perhaps. I am not sure

in redout 2 you can kind of fly around though

[cloud rivet]

good enough for now

time to work on the road/track

[cloud rivet]

TIL about __builtin_nondeterministic_value

for if you want a valid float of some type determined at compile type, though you won't have any idea what it will be

also I learned about

#pragma clang loop vectorize(enable)
#pragma clang loop interleave(enable)

there's a whole section about vectorization I should have read when I was working on my software rasterizer https://clang.llvm.org/docs/LanguageExtensions.html#vectorization-interleaving-and-predication

[cloud rivet]

I'm going to work on controller support next

and then I'll work on physics some more

[tight torrent]

beat your time

[cloud rivet]

nice

I switched to pure time

I still don't get under 50 though

I'm not good at video games

so wasn't like trying to compete lol

[tight torrent]

I know lol

Im a somewhat competitive person and I kinda wanted to play redout lol

[cloud rivet]

you can't tell, but I'm doing all this nowl via the xbox controller

[cloud rivet]

if you want to compete with me on something, compete by building a game, if you aren't already

[tight torrent]

I am sorta

https://samfizz.itch.io/biomechanica

[cloud rivet]

wow nice!

[tight torrent]

Im lead dev on this, one of my friends is the project manager and he handles the itch and steam page (when we get one)

[cloud rivet]

that looks great

[tight torrent]

Thanks

It's a FPS metroidvania with like 3-5 hours of content planned

[cloud rivet]

feels half life inspired

[tight torrent]

I see it, yeah

[cloud rivet]

this is your job?

[tight torrent]

Alot of the lore behind the game is more resident evil inspired, one of the original pitches for the game was 'what if you played as the virus that infects zombies' instead of trying to eliminate it

[tight torrent]

Nope, im a college student

Doing this on my free time

[cloud rivet]

the extra redout maps you have to pay for are really nice if you like the game

it's like $10 for all of them or there about

I really like all the maps

hrm all that I have left to do to get a full game loop is just physics work and track work, although I'm going to add debug lines so I can see the orientation of the vehicle and the turn orientation, since that has its own orientation. I just need a bunch of debug lines. I'm going to try and just do that with my software rasterizer

I'm going to get turning and banking working and then I'll work on gravity & rotational dynamics. Once that works I'll make the track a loop and have a track goal where I can record lap time. Once I have that I have a game loop. Then I'll work on making the track more interesting.

once I start on graphics, I'm going to see how far I can go on the graphics without textures just using shaders and per vertex data for colors and metallic roughness

I'd have to add libraries to get textures, but I think maybe I can make a nice looking game without textures

I'll use window's font system for game UI

[pseudo dock]

Seems like you're making nice fast progress!

[cloud rivet]

thanks! trying to keep up the pace

it's fun

my job hasn't taken up my entire day the last week, so I have had time work on this

[pseudo dock]

Oh, nice that you can get more time in on your own project

[cloud rivet]

you got me to try lol

[tight torrent]

is that a challenge?

[cloud rivet]

I'm an old man, I can't win

my favorite map is the moon map, because it has really fun jumps, but all the maps are just fantastic

once I can complete all the jumps every lap I can probably get that time way down

[tight torrent]

beat it again

[tight torrent]

i dont have the dlc, so cant try that one

[cloud rivet]

nice!

this game is so much fun I can't believe I never heard of it before

[tight torrent]

well i beat your time lol

[pseudo dock]

https://i.makeagif.com/media/8-17-2023/mDIhIM.gif

[tight torrent]

it is lots of fun

this u lol

[cloud rivet]

yes

[cloud rivet]

I actually remember playing this game as a kid

it was an arcade game in a bar my dad dragged me to when I was like 6

[pseudo dock]

I have played the Atari (2600 maybe?) and Apple ][ versions, heh, but yes, it was also an arcade game. Moon Patrol! Cool music too.

[tight torrent]

This is the Atari one

Damn

Looks like shit

But thats the 2600 for ya

[pseudo dock]

I looked it up and I definitely haven't played the Atari 2600 version (the one you show there). I might have played the Atari 5200 version, but it didn't look completely familiar to me either, so maybe my memory is wrong and I just played the Apple ][.

[tight torrent]

I remember seeing a apple ][e at an estate sale near me for 100 usd a few years ago

Other than that, im waaaaayyyy past the time that those things were relevant

[pseudo dock]

They're still relevant if you want to learn some history by playing some old classics 🙂

(But I understand what you mean)

[tight torrent]

Yeah, I think my dad has some floppy disks with some old programs

No clue if they'd work

But if I had an apple ][

Might give em a shot

[wraith urchin]

What is the music?

[cloud rivet]

Generated with Suno.

It’s not in the game

I was just listening to it

My game doesn’t have audio

If I made a real game I would pay someone for real music

[cloud rivet]

You know redout has a soundtrack they sell on steam

I should listen to that instead

I really like the music in the game

[cloud rivet]

No steering yet, but I added debug lines and am handling left thumb controls now

[cloud rivet]

got a bunch of stuff to do today

I don't know if I will make much progress

I have steering and bank controls input, now I just need to make it work with the vehicle

it's just a bunch of physics work to do now

[tight torrent]

i think youve got me beat on alaska speedway, cant crack 45

[cloud rivet]

How do you like the moon map?

It’s so much fun

[tight torrent]

its pretty fun

i still technically dont own the dlc, but the content is in the game still so you can spoof the calls to epic or steam to return that the dlc can be played lol

[cloud rivet]

lol

[pseudo dock]

Buy it! You'll want others to pay for your game once you release your game (or eventually once you get a job in the industry)

[tight torrent]

yeah i will once i get the money for it

[cloud rivet]

ok I think I got a good idea of semi-implicit euler, getting total forces and torque and getting orientation, like high level understanding, the book I am using for physics (https://www.essentialmath.com/book.htm) has a bunch of demo code I forked, and I'm going base my physics off it

https://github.com/btipling/essentialmath/tree/master/src/Examples/Ch13-Simulation

it's great, book and code all in one, gamemath.com also has physics, Eric Lengyel hasn't published his physics book yet, but his C4 engine has physics which I licensed

I briefly skimmed physics in a weekend and it's not really what I want to do

this going to be some super basic stuff I'll have to iterate on, it's not a real world simulation, it's gonna be rough and I'll work on making it better over time

[pseudo dock]

You probably know of these, but since you're listing sources and it seems like you have been looking...

I have the first edition of this: https://www.amazon.com/Game-Physics-Engine-Development-Commercial-Grade/dp/0123819768/
I'm not a physics programmer, but it seems ok 🤷 . It's old enough now that it might very well be outdated but the author takes a very practical approach based on his actual development experience which I like.

For collision detection I have this:
https://www.amazon.com/Real-Time-Collision-Detection-Interactive-Technology/dp/1558607323
I hadn't anticipated it when I bought it all those years ago but I have ended up referencing this book many times for graphics and just efficient programming for real-time in general. Even though it's old (and apparently never got a second edition?) it has ended up being one of my favorite books from back when I bought a bunch as a student.

(I also have that the first edition of that essential math book and have found it to be a pretty good reference. Do you have a newer version? I wonder if I should buy an updated copy one of these days.)

[cloud rivet]

I have had this on my to buy list, but I haven't bought it yet. I haven't heard of the second one. Thanks!

I rolled my own physics in my blockens voxel game, I just did a bunch of ray casting and had nothing resembling real physics, but it worked

for walking around and falling and not running into things

I think with this antigrav thing I think I can get away with something less than realistic too

I think I will eventually get more into it depending on how long I end up living lol

going to add these books to my read list thanks again

[pseudo dock]

For my block game I've been working on I decided to do all the physics because I thought it would be interesting to actually implement rather than just reading about it and the game seemed simple enough with axis-aligned boxes with no rotation and a sphere and that I could actually do it. (If I were doing a real game I would probably just use Jolt, but I thought I should at least get some experience actually doing it myself.)

What I learned:

• Getting all the corner cases to work right because of numerical imprecision is hard. I "knew" this before trying because I had heard it, and I totally believed it, but actually experiencing it was pretty interesting. I think there is still a case where my code can get stuck in an infinite loop but I don't know how to reproduce it in order to try and diagnose the problem and fix it hehe.
• Dealing with acceleration is much harder than I would have guessed. If it's just velocity and position things aren't so bad, but dealing with acceleration with discrete timesteps has been a much bigger challenge than I would have anticipated.

I'm excited to see what you come up with! It seems like your game has more complex physics than my simple one which would intimidate me but you seem to be making pretty quick progress.

[cloud rivet]

Maybe tests might help with those edge cases?

Your game is very cool

[pseudo dock]

Tests? What are those? 🤓

[astral hinge]

code that transforms your code into corporate code

[cloud rivet]

I think I will use hot reloaded lua files to design my tracks

and avoid any kind of editor UI

code will be my UI

that means I'll have to add my first dependency, Lua

and I'll just edit it in neovim, and upon saving the file it just reloads in my running game

neovim support for lua is super good

I am hoping I could maybe use some other app to design the tracks and get the curves from

[broken fog]

you can just import bezier curves from svg

that would be 2d only, though

[cloud rivet]

yeah that wouldn't work for me

[tight torrent]

blender has splines, maybe creating a simple spline importer/exporter for blender could work?

plus then if you wanted to model an enviornment or something the track is right there in blender, which might make it easier to model around

[cloud rivet]

yeah I was thinking of using Maya or Houdini, but feels like a lot

[cloud rivet]

these controls need a lot of work, but I have something to start with. I converted all my orientations to quaternions also. I will work on gravity and torque next

actually I'll work on making the track a loop first

[cloud rivet]

ok gravity is next

I have learned it's important to show a little bit of the turning the vehicle is doing in 3rd person too, being stuck right behind the vehicle in the exact same orientation is nauseating a tiny bit and looks strange

[astral hinge]

nice, is that collision I see?

or are you just stopping the car when its altitude is below a value

[cloud rivet]

no I don't have any collisions yet

the vehicle stops when it is not hovering

I wish I was smarter

I think I lose brain cells at work

[cloud rivet]

I have to work all weekend, so going to lose some more brain cells instead of making progress

[cloud rivet]

I got a lot done, maybe I have some time tomorrow

[cloud rivet]

reading the redout and redout2 reviews from really skilled and knowledgable players is really meaningful way to learn what they like about the games and what mechanics are important

redout2 performs very badly for me, it stutters quite a bit even though I have more than the recommended specs. given the glowing reviews I wonder what's wrong

I have every other user application shut off and even on medium settings there's some stuttering

I think redout looks better visually than redout2

even just the cinematics stutter

it's so weird

I don't have any problem like this with other games

[gaunt drum]

vroom vroom

[cloud rivet]

had to work all of today too

[broken fog]

why do you have to work on weekends

hope you're getting paid for that overtime, at least

[solid grove]

the joys of being salaried

[cloud rivet]

no I don't get overtime

nobody asked me to work weekend

I just had to

It was important work

[wraith urchin]

important for who?

[cloud rivet]

For the company, for the customers who put their trust in us, for my team, for the week ahead,

[hushed creek]

that's such a bjorn thing to say, love it

[astral hinge]

for my 401k

https://tenor.com/view/aragorn-frodo-run-gif-12983861

[broken fog]

i mean, it's good that you care but also look out for yourself

[cloud rivet]

Yes

I was the second employee hired at the company fwiw. I started a month after the company was incorporated. I care a bit more because of that I guess

I am happy to get back to my game today

Was hoping to have made a lot of progress on the track physics

[astral hinge]

is the founder the first or zeroth person hired?

[cloud rivet]

well if we count founders I'm 4th

I think I accepted my offer before the other employee though, my start date is just after them :P

we're hiring btw https://www.conductorone.com/careers/

pls help

[broken fog]

fair enough, the way you talk about it on here made it sound like you were being made to work on weekends

if it's self inflicted all i can say is remember to take time for yourself as well and don't be exploited, take care my frog

[cloud rivet]

well I'm not happy about it

I wish I had more help

[wraith urchin]

Do you have equity in the company?

[cloud rivet]

yes

[hushed creek]

Thats good at least

Good reason to work harder assuming you can cash in at some point

[bronze socket]

what do you guys do?

[cloud rivet]

access requests and access reviews and access automation

[bronze socket]

ah, it looks like the vast majority of what you have open is frontend stuff

I don't even know what a connector is

[cloud rivet]

no

there's only one frontend role in that list

a connector is an integration where you pull data out of a service like aws

and then we bring that data into our platform

most of these positions are Go backend fullstack positions

[bronze socket]

so it's like an ETL?

[cloud rivet]

I don't know what an ETL is

oh sort of yes

only single source

[cloud rivet]

https://www.forbes.com/sites/aliciapark/2025/10/28/ai-employees-are-coming-this-startup-wants-to-keep-them-secure/

we raised $80m for our series B

Hopefully I can get help now

Anyway, I actually managed to get a couple of fixes into my game today

I didn’t know about ‘saturate’ in HLSL

That’s neat

clamps to 0 to 1

Available in slang too

[slim oak]

yeah, it's nicer than having to type clamp(x, 0, 1);

[cloud rivet]

I should make progress on the track but the way the camera follows the vehicle bugs me

I just hard set it to follow directly behind the vehicle with the same orientation and it is so weird

What I am going to instead is have the camera follow the path of the vehicle instead

Like it’s the vehicle a few moments behind

I think maybe I can just get away with a sequence of lerps but maybe I need to track the the vehicle path, which could be reusable for replays too

[solid grove]

my camera lerps to the player position and the root object lerps to copy the rotation of the kart

[cloud rivet]

nice, that's what I was think I would try

what is the root object?

I just have a camera

it has a position and orientation

[solid grove]

my camera is a 3-object transform hierarchy. the root object, the boom, and the camera. The root object controls the position and rotation, the boom specifies the offset from the root, and the camera object allows for some additional tilt and offsets

[bronze socket]

what made you pick that solution

that sounds like a slightly more complex way of specifying the camera as offset by a translation and some euler angles from its mounting point

[solid grove]

because it doesn’t require any thought and it’s easy to add animations to

eg the camera changes what it does during the blowing up animation

if i wanted to add all the things I need to do with the manual offset solution id end up making a form of transform hierarchy anyway

[cloud rivet]

nice

[solid grove]

i used a very similar setup for animating cameras in the offline animation world

[cloud rivet]

I have seen your YT videos, very cool

[solid grove]

lol those are so old 😆

[cloud rivet]

not at my age they aren't

[cloud rivet]

just working on the camera a bit

I like it, I kind of want it to have a greater minimum distance from the viewer, I'll work that out later. I really want to work on the track now

I need to slerp the orientation change

that's better, but something is still shaky, I am really going to work on track physics next and worry about it later

[cloud rivet]

been hard to make any progress on anything this week. but I hope to this evening get back to normal

this is an ad that our customer, Ramp, paid for to thank us?

[tight torrent]

were you visiting NYC?

[cloud rivet]

no

we have some people who work in NYC and they took these photos

[tight torrent]

ah

[cloud rivet]

and shared with everyone on our company slack

the first is NASDAQ tower, and the second is on the NYSE stock exchange, I don't know where the billboard is that Ramp paid, I think NYC Time Square

it's been a crazy week

oh it says right in the picture where it is

[cloud rivet]

yes

[tight torrent]

ah gotcha, what was the big event for your company?

[cloud rivet]

we raised our series B

we convinced a bunch of well capitalized VCs to put their hopes and dreams in us

and got enough cash to go big

[tight torrent]

very cool

[cloud rivet]

yeah, hopefullly we dont screw it up

[tight torrent]

but i imagine it means everyones very busy lol

[cloud rivet]

unbelievably busy

I have worked for two weeks straight at this point, into the late evenings

[tight torrent]

yikes

[bronze socket]

the new york NYSE stock exchange of course

[cloud rivet]

lol

I'm good at english

nearlly as good at it as I am at spreading vulkan sync

[cloud rivet]

first attempt at trying to use Houdini to make some points for a 3D curve

a tiny bit more interesting track

and then I'll use python in Houdini to generate a C file with the points

and eventually I'll hopefully get better with Houdini to make the tracks more interesting

pretty much exactly what I wanted, a way to visualize the curves

[solid grove]

first indie dev I've seen use houdini

do you have a vfx background?

[cloud rivet]

no :/

[solid grove]

what was the reason for choosing houdini?

[cloud rivet]

I watched a talk from the Ascent devs and they said they used houdini

and I saw it had an indie license

I can export my points via a python sop node

so I think I can write a file with python directly to my project

[solid grove]

i do that but with blender

I have a bpy script that outputs a binary files as a build artifact for the game that contains track data

[cloud rivet]

nice

I have a full flipped normals houdini course I haven't gone through yet, I was planning on doing

I took multiple courses for Maya and Marmoset and it was really great

so I think I can end up with some nice tracks, and then probably reuse the curves to actually generate geometry for the scenery around the track

the stuff you can generate with houdini is pretty remarkable

[solid grove]

I don't use curves to make the track collision mesh, i model that by hand and export a gltf for it. but I do use curves to place the track checkpoint marker planes

[cloud rivet]

your track is great

as are your animations

I don't know what I'll end up with

I'm kind of just exploring the tool

maybe it will be garbage

[solid grove]

https://cdn.discordapp.com/attachments/493896891562721280/1376418654358339706/checkpoints.mp4?ex=6904e583&is=69039403&hm=b7a3f7ebf777e92915e8db83a0e966351e99036fbc45b63cbbe97bf8018438f9&

placing all of those by hand would be terrible

[cloud rivet]

oh interesting

ok so Maya has a similar curves tool as that

I have used it to do animations

and ropes

[solid grove]

this is a blender modifier that copies a quad mesh over the curve in fixed increments. then I have a bpy script that iterates all of the faces of this new mesh and writes out their center and normal so that the game can load them in as trigger colliders

[cloud rivet]

what are trigger colliders?

[solid grove]

colliders that don't block objects that touch them

but still register collision events

[cloud rivet]

what do you do with those events?

[solid grove]

when karts pass through a checkpoint plane collider that's how I know where it is on the track

I use it for determining what place (1st, 2nd, 3rd etc) each kart is in, and the AI use it to follow the track

[cloud rivet]

oh that's smart, is that something you read about?

[solid grove]

yeah its a pretty typical technique in racing games

all of the colliders have numeric IDs and the karts store what the last checkpoint ID they touched was, along with the game time that they touched it so that if two karts are between the same 2 checkpoints they don't both show as being in the same place

this also lets the game know if a kart is going the wrong way, as well as when laps are completed

[cloud rivet]

this is great, I think that will be useful for me too, thank you

I am not sure I will have an actual race with AI or anything, I was just thinking of making a simple beat your previous time game play. I don't even know if I can do the physics or how far I will get. I'm taking one small step at a time. I have a really basic curve I just want to now write into points to use to generate the mesh in the mesh shader, and then I'll also have the points in the app and can use that to determine the plane that will serve as the floor and emits the normal force so I can have gravity

[solid grove]

mario kart 8 also uses this mechanism to know when you fall out of bounds. the planes can be connected to form a tube around the track and if a kart ever leaves that tube then the game knows that the kart is out of bounds and lakitu should return it to the track

i'm not going to do that and just opt for a normal kill plane since the tube is a bit more complicated

[cloud rivet]

what's a kill plane?

[solid grove]

if kart y position < some value then you fell off the track, trigger the rescue system

[cloud rivet]

I think a tube makes sense. for my game I was thinking for the anti grav mechanism, the normal force is actually an attractor and if you're within right side of the plane and some distance that's where gravity is pushing you to

not sure yet

next step is just turning these points into C

[solid grove]

the tube makes sense for games with antigrav, older mario kart games also used a kill plane since the tracks were flat

i think f-zero also uses a tube

[cloud rivet]

pretty cool I can add a button to export the track with the python script

the 4th value is the curveu value

C:\Users\Bjorn\projects\code\fury_run>type src\p_track_1.c
// Auto-generated track data from Houdini
// Points: 50
// Length: 0.80

const float track_points[][4] = {
    {0.1220f, 0.0829f, 0.0000f, 0.0000f},
    {0.1210f, 0.0716f, 0.0156f, 0.0204f},
    {0.1180f, 0.0791f, 0.0309f, 0.0408f},
    {0.1131f, 0.0856f, 0.0458f, 0.0612f},
    {0.1063f, 0.0892f, 0.0599f, 0.0816f},
    {0.0978f, 0.0855f, 0.0730f, 0.1020f},
    {0.0876f, 0.0750f, 0.0849f, 0.1224f},
    {0.0761f, 0.0645f, 0.0954f, 0.1429f},
    {0.0632f, 0.0570f, 0.1043f, 0.1633f},
    {0.0494f, 0.0531f, 0.1116f, 0.1837f},
    {0.0347f, 0.0531f, 0.1170f, 0.2041f},
    {0.0195f, 0.0531f, 0.1204f, 0.2245f},
    {0.0039f, 0.0531f, 0.1219f, 0.2449f},
    {-0.0117f, 0.0531f, 0.1214f, 0.2653f},
    {-0.0271f, 0.0531f, 0.1189f, 0.2857f},
    {-0.0421f, 0.0531f, 0.1145f, 0.3061f},
    {-0.0564f, 0.0542f, 0.1082f, 0.3265f},
    {-0.0698f, 0.0604f, 0.1001f, 0.3469f},
    {-0.0820f, 0.0694f, 0.0903f, 0.3673f},
    {-0.0929f, 0.0807f, 0.0791f, 0.3878f},
    {-0.1022f, 0.0884f, 0.0666f, 0.4082f},
    {-0.1099f, 0.0881f, 0.0529f, 0.4286f},
    {-0.1158f, 0.0800f, 0.0384f, 0.4490f},
    {-0.1198f, 0.0753f, 0.0233f, 0.4694f},
    {-0.1217f, 0.0717f, 0.0078f, 0.4898f},
    {-0.1217f, 0.0829f, -0.0078f, 0.5102f},
    {-0.1198f, 0.0829f, -0.0233f, 0.5306f},
    {-0.1158f, 0.0829f, -0.0384f, 0.5510f},
    {-0.1099f, 0.0829f, -0.0529f, 0.5714f},
    {-0.1022f, 0.0829f, -0.0666f, 0.5918f},
    {-0.0929f, 0.0829f, -0.0791f, 0.6122f},
    {-0.0820f, 0.0829f, -0.0903f, 0.6327f},
    {-0.0698f, 0.0829f, -0.1001f, 0.6531f},
    {-0.0564f, 0.0829f, -0.1082f, 0.6735f},
    {-0.0421f, 0.0829f, -0.1145f, 0.6939f},
    {-0.0271f, 0.0829f, -0.1189f, 0.7143f},
    {-0.0117f, 0.0829f, -0.1214f, 0.7347f},
    {0.0039f, 0.0829f, -0.1219f, 0.7551f},
    {0.0195f, 0.0829f, -0.1204f, 0.7755f},
    {0.0347f, 0.0829f, -0.1170f, 0.7959f},
    {0.0494f, 0.0829f, -0.1116f, 0.8163f},
    {0.0632f, 0.0829f, -0.1043f, 0.8367f},
    {0.0761f, 0.0829f, -0.0954f, 0.8571f},
    {0.0876f, 0.0829f, -0.0849f, 0.8776f},
    {0.0978f, 0.0829f, -0.0730f, 0.8980f},
    {0.1063f, 0.0829f, -0.0599f, 0.9184f},
    {0.1131f, 0.0829f, -0.0458f, 0.9388f},
    {0.1180f, 0.0829f, -0.0309f, 0.9592f},
    {0.1210f, 0.0829f, -0.0156f, 0.9796f},
    {0.1220f, 0.0829f, 0.0000f, 1.0000f},
};

const int track_point_count = 50;

these values don't match my current scale

pretty cool, I'll figure out how to use this to send to my mesh shader

then I will figure out how to use it for a floor

I'll do that tomorrow

MJP has a whole series on sync https://therealmjp.github.io/posts/breaking-down-barriers-part-2-synchronizing-gpu-threads/

[cloud rivet]

I don’t want a c file I want to be able to reload this that’s right

[astral hinge]

so will you use csv or something?

[cloud rivet]

Yes something easy to parse

[bronze socket]

like a fixed width EBCDIC file, fresh from the mainframe?

[cloud rivet]

I was just thinking to use scanf 😅

https://learn.microsoft.com/en-us/cpp/c-runtime-library/reference/sscanf-s-sscanf-s-l-swscanf-s-swscanf-s-l?view=msvc-170

sscanf I mean

[solid grove]

my format is even dumber than that. I write out the data as binary and then memcpy it into an array of structs

no parsing needed

[cloud rivet]

I did that with Rosy

it worked

but I'll parse some ascii for now

[cloud rivet]

ok I have my points in my application and now I need to make them available to the mesh shader and use them to generate the track

[solid grove]

do u need to generate the track in a mesh shader?

[cloud rivet]

I just want to use a mesh shader for some reason, I don't have to

I'm trying learn how to use mesh shaders, figure out how they're valuable or useful

I thought for a track it made sense

still think so

I don't want to load in an obj or gltf for the track

[tight torrent]

their fun

[cloud rivet]

it's definitely been so far

I think I will generate the track and its barriers in the mesh shader, the vehicle and things that occlude the track will be via obj files that I'll make in Maya or Houdini, and then the rest of the scenery I want to RT

[tight torrent]

RT is fun

way easier than raster stuff imo

[cloud rivet]

I hope so

the RT stuff will just be distant static scenery with no depth buffer

basically the background

[solid grove]

particles are pain with RT. I still haven't figured out how i'm going to restore particle shadows yet now that I don't do shadowmapping anymore

[cloud rivet]

how did you do that with traditional shadows?

[solid grove]

just render the particles again to the shadowmap

it was basically free

[cloud rivet]

right, but you'd have to update the blas/tlas to do that with RT

so that's not free

[solid grove]

plus particles are even faker than regular game objects. the sprite-based ones are billboarded quads and that concept doesn't exist in RT

so I'd have to use the AABB BLAS mode and calculate the bilboard inside there on intersection somehow

[cloud rivet]

are shadows for particles common?

[solid grove]

most games don't bother, but it looks nice so I want to do it

really helps ground the effect in the world

[cloud rivet]

might not be too expensive to just keep using a shadowmap?

just for those

[solid grove]

well I deleted all that code because it only semi-worked in the first place

i think for the initial vertical slice particle shadows aren't happening

[cloud rivet]

what kind of particle casts a shadow?

[solid grove]

smoke and item box shatter fragments

u can kinda see it in this old recording https://cdn.discordapp.com/attachments/1173035001554677863/1411489105274994828/HKart_2025-08-30_16-11-36_-_Trim.mp4?ex=6906942a&is=690542aa&hm=1cc2cde0cc3de9e39ce2c9ed493cb8b1f647a07ddee77a613fb18b3122f48f5f&

[cloud rivet]

interesting, wonder how you'll solve that

[cloud rivet]

looks pretty good at the end yeah makes sense

hrm

[solid grove]

that weird dithering on the fragments is the video encode i'm pretty sure

I don't have any dithering in my code

[cloud rivet]

I believe it

I see weird stuff in my videos too

I remember trying to debug something I saw in a video

then when I thought I figured it out it wasn't even real in the game

in my softwarer rasterizer a few months ago

it had like a weird magenta line around one of the cubes in the video

[cloud rivet]

isn't there a variant of one of the RT shader where you just invent geometry?

I forget what it's called

it's like programatic, I didn't look into it much

looking it up now

oh

I think it's a combination of the acceleration structure mode and VK_KHR_ray_tracing_position_fetch maybe

[solid grove]

possibly, its definitely interesting

thanks for finding that

[cloud rivet]

like VK_GEOMETRY_TYPE_AABBS_KHR

https://github.com/nvpro-samples/vk_raytracing_tutorial_KHR/blob/master/ray_tracing_intersection/README.md

that might be an approach or useful in the research for it

it doesn't need actual vertex data

[solid grove]

yeah my plan is to represent the particles as AABBs and then try to reconstruct the billboarded quad in the intersection shader

[cloud rivet]

track data is on the gpu, now I need to figure out how to use it :O

the scale is still too small too

I'll just wing it in the shader for now?

once I draw this track it will immediately break the illusion that I have any physics since the track has elevation changes

final video with the fake physics before I dump it, I do like my third party camera, it does need a bunch of work but it's fun

[cloud rivet]

so just have been thinking about what I'll do, I'll take the vector from one point to the next to use as a plane normal, the point is a point in that plane, and do a couple of cross products to get two orthogonal vectors, I'll treat one as a line, I'll use the same line for the next point to draw a quad. The plane will serve as the same thing as what ravbug explained. Since I only have 50 points per track, which I think is enough to describe the track, but not enough to render, I'll lerp some intermediary points to smooth out the track.

that's just a theory of what I'll do, I'll figure out how that works out when I start writing code

I'll need to eventually generate a normal per point in Houdini to server as the normal force, since I want the track to have twists and loops

I'll just assume y up to get the initial track and physics working

[cloud rivet]

I tried looking into OpenAI codex as an alternative to claude code and what a shit show. I can't believe how hard it is to create an windows agenting workflow with Azure on the CLI wheras it's as simple with anthropic as an npm install and logging in with claude. You have request forms, sign with openai into azure, you have to use WSL. I can't believe how hard MS is dropping the ball here. Their product in 2025 is just a shitty auto complete, the same as it was in 2023.

everything outside of anthropic treats windows as a second class citizen too. I just don't get it. Feels I just want to quit my job right now and start a startup to fill this gap or something

just like a cmd/pwsh first CLI agenting workflow thing, windows/azure dedicated

how is this not a thing yet

[solid grove]

seems like the AI field is composed mostly of people using Macs to remote into servers running Linux. anecdotally of my friends who work at AI startups, for all of them the company issued them macs

There’s stuff made by Microsoft Research that runs on Mac and Linux but not windows

[cloud rivet]

yeah same at my job, but at ConductorOne we have customers who are full windows shops. Like I ask them to do a curl to test something and they do it in powershell instead

people are sleeping on this

I'm talking about utility companies, banks, manufacturers

they aren't in the SF macbook bubble, they use active directory and just have C# apps and oracle and microsoft licenses

like real companies that do things, not vc funded tech dreams that do nothing

[solid grove]

does Claude code work on windows natively?

[cloud rivet]

it does

it's great too

but you're forced to use anthropic models

which seem to be falling behind

sonnet 4.5 is very bad, an opus 4.1 is really expensive token wise and anthropic is calling it "legacy"

I am not sure anthropic is able to keep up with openai, I don't really understand what's going on with anthropic which has been leading imo up until recently

[solid grove]

that industry is a revolving door so maybe one of their top people got poached

[cloud rivet]

I don't know how to build an agentic workflow, I just want to work on my game

I don't use ai anyway for working on it

[solid grove]

for one of the startups a friend works at, the startup got a Facebook person for some ungodly sum, but that person only stayed for 2 weeks before hopping to another startup for an even bigger sum

[cloud rivet]

what an obvious bubble

I'm going to try and ride it out at my current job, we just got enough money to last a few years I think we'll be insulated from when all this fallsapart I hope

idk what I'm going to do

work on my game for now

[astral hinge]

what's next on your list

[cloud rivet]

drawing the track in the mesh shader with this approach #1323084490997895198 message

[cloud rivet]

ok

uh

well

I guess the vehicle has to go on the track now 😅

maybe if I ask it nicely

how do I put this vehicle on the track? wrong answers only

[astral hinge]

trebuchet

[cloud rivet]

ok ok time for physics, I will drop it on one of the points and it will fall until it hits the track

[astral hinge]

https://youtu.be/NLiL0GLSvIw

[cloud rivet]

that's amazing

TIL https://www.sidefx.com/docs/houdini/nodes/dop/bulletrbdsolver.html

it has bullet physics nodes

[astral hinge]

seems weird to leak the underlying physics library's name like that

[solid grove]

maya does the same thing with its bullet solver
https://help.autodesk.com/view/MAYAUL/2026/ENU/?guid=GUID-325F6D2C-C0AD-4A6D-9F37-970CE7F32FE4

but it has multiple solvers (nDynamics is the default)

[tight torrent]

rocket

[cloud rivet]

I think I can avoid writing actualy physics to solve this problem by deleting my project

jk

[tight torrent]

have the track be 2 splines, one for a inner and outer edge, then make it so if you get too close to the edge you bonk the wall

then you can do like banks in the track with just raising or lowering one side of the track

[tight torrent]

if you want a distraction, redout has a VR mode

(i wanted an excuse to talk about redout vr, did not know that exists)

[cloud rivet]

ah interesting, I don't own a vr thing

[cloud rivet]

ok, I think I understand more about the physics, how to apply the force of gravity to each point on the vehicle both for linear dynamics and for torque, the linear dynamics cause the change in direction, the torque contributes to the orientation, eventually the vehicle could rotate enough where none of the points were colliding with the ground and be in free fall. upon contact for any of the points of the vehicle I apply the normal force which should cancel out the gravitiational force. the torque for each point is a cross product of the position of the local point and of the force being applied to it. I can apply a center of mass for the vehicle to each point to cause the orientation to be centered there. Using implicit Euler, so I get the linear force, which involves the vehicles velocity and gravitational force, apply a drag/friction calculate the acceleration which updates the velocity which is then applied to the position. Angular motion is taking the torque applied to each position with a cross product, taking into account center of mass, this becomes angular velocity, I add all these torque forces for each point into a total torque force, I calculate an axis angle by taking the normal of the angular velocity vector as the axis of orientation and use its length as the angle, and that I convert into a quaternion delta I multiply by the vehicles orientation

I think I can do this

I will try tomorrow

[brisk chasm]

i read "i do like my turd" 😄

[cloud rivet]

I currently have a bunch of stubbed out methods and set up some new structs. I have the basics of the integration function written using the stubbed out methods. I just have to write out the actual code of the functions and then I should have physics

and then just make it better

once I get the very basics of physics working I'll write an obj loader and create a little bit better vehicle in Maya

[cloud rivet]

I have to go to portland tomorrow for the week for work so won't be able to make progress after today until friday

I'm going to focus on starting on filling these structs with functionality I can actually see happening instead of trying to get it right all at once blind

[brisk chasm]

say hi to cyberreality

[cloud rivet]

just there for work stuff

[brisk chasm]

i wasnt serious

[cloud rivet]

so this is where I'm dropping the vehicle in, ok let me start with gravitational force

[cloud rivet]

I don't want to get sucked into the matrix

[cloud rivet]

ok gravitational force

going to add some random torque to see how it behaves

then I'll work on track collision

[cloud rivet]

fun

ok collision time

the collisions are per vehicle contact point, which also contribute to torque, so it should enable the vehicle to move up inclines and come down declines

and look correct

there's some shenangians with that since there are no contact points the thing hovers

but it hovers with 4 rays right now from each corner

any ray that's not penetrating the track gets the full gravitational force applied to the torque on that point, curious how that ends up behaving

I don't have center of mass yet

I also don't have any slope parallel force yet

I just want a basic thing to work, the vehicle stays on the track or falls off, I'll make it better later

[cloud rivet]

so for collisions

each point in the track currently has an assumed up roughtly equal to y up

I take a track point and the next and create a quad from these

I create 3D bounds object for the 4 points this quad has

for each contact point on the vehicle I detect if it's not within the bounds object (a bounding box)

in that case not in contact with the floor

if it is within the bounds

I calculate the actual up vector of the plane the 4 points make by taking a cross product of the right and left vertices (eventually I'll generate the up vector in houdini, as I need that to create loops and twists in the track)

and then I use that up vector as the surface normal for a plane and for each point determine its distance to that plane, if it's negative it's not in contact with the ground, and if it's within a small distance it's hovering

that'll produce upward torque, and the normal force on that point up until the maximum hover point

that should keep the vehicle on the track I think

that'll be a per vehicle contact point per track point so 4 x 50

I think that's fine

[cloud rivet]

lol

[cloud rivet]

ok well

I think I need to read more

[cloud rivet]

so part of the problem is my code was just finds the first part of the track it's on, and this vehicle is dropping on a spot where it covers two so in the first video for two contact points it doesn't see a floor and is trying to fall but can't, so there's a couple of bugs there, where I need to capture all the parts of the track that I'm on.

in the second video I removed applying the gravity for torque to simplify that code and just get it to be stable so I could fix one thing at a time and I applied a hover thrust to my linear force pushing up with an incorrect value and logic. And then in the end there's a bug where it's not applying the player's velocity in the direction the track is going so it goes through the track, and there's bug where it doesn't fall, it just drops to Y 0, and keeps going.

it's just a lot of bugs

I was rushing through this trying to get something working before going to portland

I spent most of the time just setting up the scaffolding to even have physics. Now that I have it I can spend more time on fixing the bugs and getting the parameters & logic right

[pseudo dock]

Hopefully you will be able to concentrate in Portland rather than being distracted thinking about your physics waiting for you at home 😅

[cloud rivet]

I would rather think about physics for my game than work. Working on my game is how I keep from burning out

It’s the only thing that I get meaningful amount of serotonin from outside of exercise

The rest of everything else in my life is just endless responsibilities

Which I am thankful to have but takes up so much energy

If I didn’t have this project and exercise I would be an empty shell

[pseudo dock]

Don't you dare go hollow!

[cloud rivet]

ok I am back today

I figured out some of my issues from sunday, I hadn't removed all of the fake physics code I had from before

and it was conflicting with my new physics

I disabled torque, so it looks funny, but removing the old fake physics I had before removed some of the weird stuff

I thought I had removed all that code but I hadn't actually

[cloud rivet]

so torque actually works, except for two bugs

one is my hover code, before I added real physics I had hard coded hovering, and I haven't yet done all the work to support it, that's why I think the vehicle is kind of angled upward

the other bug is that my torque logic is based on finding the track quad beneath the vehicle and isn't able to handle being across two quads

so I'll fix that

and then I'll fix the hover code

and then I think I should be close to being able to go around the track

working on this bug is making me physically nauseated tbh ugh

[pseudo dock]

I guess it's good you're working out all the bugs before making the VR version

[cloud rivet]

I'm going to need to track a bunch of state per vehicle contact point

I'm too tired to do it now, I have to sleep

I wrote a big note to myself in my editor in the physics code so I don't forget

[cloud rivet]

I fixed a lot clipping bugs, still pretty buggy at steep slopes,

the controls are hilariously difficult atm

[brisk chasm]

perhaps your soap box needs some whiskers

to detect borders like a lidar

[cloud rivet]

it needs something

right now I do a check to see if a new position would penetrate the surface and that mostly works

[brisk chasm]

thats what i mean

you might need more checks (like more whiskers) to do that, around the front of the box

[cloud rivet]

yeah maybe, I am just checking 4 points, although I think I have an incorrect slope normal somewhere or something

[brisk chasm]

perhaps a debug viewer for slope normals is the next thing then

[cloud rivet]

oh debug views hrm

good ideas thank you

I think I should add some points in the front of the vehicle for sure

[brisk chasm]

like a hemisphere or like those rams on those old steam locomotives

[cloud rivet]

yes that makes sense

[brisk chasm]

https://tenor.com/view/buster-keaton-hat-surprised-surprise-shock-gif-7670153191211300805

https://tenor.com/view/chaplin-train-chaplin-gif-26687482

[cloud rivet]

these controls are kind of fun even though they're terrible

it's really hard to stay on the track

I'm trying to fix bugs but I just find myself trying to stay on the track

[brisk chasm]

maybe you add 2 skirts around the track with a slightly higher slope

so that you, when you get off track, dont immediately fall off, but have time to correct the course to get on track again

[cloud rivet]

yeah the goal is to have rails, redout sections without a rail are also very difficult

[brisk chasm]

ah so different "layers" of skirts so to speak, the morer you are off track the more the game will tell you to get back on track 🙂

[cloud rivet]

well the sides will slow you down with friction

[brisk chasm]

yeah

[cloud rivet]

I gotta solve that falling into the track business before all that I think

I'll try some more points and then I'm going to add debug views

[brisk chasm]

sounds like a plan

[cloud rivet]

I think making the track wider and vehicle easier to control will help me debug my slope collision issues

Right now it feels like a GTA challenge to just get to reproduce it

[cloud rivet]

oh

[cloud rivet]

I have added debug lines for my rays

I'm not using the rays yet

but I have lines

now I will start using them

I'll start by making them change colors when they detect a collision

and then I'll use it to prevent collisions

I'll use the rays instead of the actual contact points on the vehicle

having the rays change colors will be really helpful because it's not obvious to me why my current collisions don't always work when they should

[cloud rivet]

amazing

so I have them colored red when they detect a collision

and that's clearly incorrect

I'm dropping it directly on the problematic sloped track quad now

it's not taking the orientation into account

it's just totally incorrect

this explains stuff

I need to add visualizations for the track normals next

I'll probably add a visualization for which track quad is being contacted too

I'm going to add visualizations until I understand every piece of what's going on

[solid grove]

you now have the same debug UI capabilities that I had for the first 5 years of development on my engine

[cloud rivet]

it's been a struggle and I keep just finding myself trying to guess at what the problem might be, but I need to stop that and just get data, I'm wasting a lot of time on hunches that aren't helping me fix the issue. There's a lot of bugs and they are sort of exacerbating each other, and I need to find just one thing that's not behaving as I expect and troubleshoot it, and that's the collisions code.

I turned off torque and other things

[solid grove]

i approximate my karts as spheres and only use the track normal for orienting the mesh for display purposes which makes it a lot simpler

i think f-zero does something similar

[cloud rivet]

well, my game is a bit different as it is an antigrav game so I can't assume gravitational force has a constant direction, I need the track normal for that

the other thing I'm doing that's making this extra complex is just using a series of points to define the track

I am actually currently assuming gravity is pointing down to simplify things

[astral hinge]

how will you support multiple planets though?

jk

[cloud rivet]

one of the redout2 maps has a black hole

it's pretty cool

[solid grove]

i think u can calculate the closest point on a spline to a given point in 3D space, then use that point's normal as the gravity vector

[cloud rivet]

yes

well right now I just have buggy collision code so I am going to draw some track surface normals, tomorrow though as I'm kinda done for the day now

I'm gonna listen to a season 2 episode of Critical Role while I go fold my laundry and then go to bed 🇬🇳

[cloud rivet]

now it's making more sense

working on track normals now

[solid grove]

me driving to work: man i wish driving was more interesting
the humble road ramp:

[cloud rivet]

at least I have a visual representation of when collision is happening, now i need to know why it fails, but having a visual track normal will be really helpful

it's not meant to be oriented this way

I turned off torque, it should straighten out and be parallel to the track and hover above it

but there's just too many bugs to get that to work right now

well

that explains it

some of it

it's not able to detect that it's on two track points at a time

and it also gets into a bad state where it fails to detect collisions at all

[brisk chasm]

bjorn pivoting to a tony hawk clone already 🙂

[cloud rivet]

holy fuck

I hadn't compiled my track shader in forever

my track was rendering wider than it actually was

it was actually this wide based on the math

[solid grove]

gotta set up automatic rebuilds for your shaders

[cloud rivet]

jfc

I lost so much time on this

well

cool cool

well I did build some nice debug stuff because of all this

I removed so much code too, I have to start adding it back in

I removed torque, slowed everything down, did all kinds of things until I finally opened up the points in a text file, added debug ui showing the points, stepped in a debugger and finally looked in render doc and started debugging slang, when I compiled it it shrank the track to the correct size

I do need like a collision pill or something eventually

just trying to get to being able to drive on the track 😭

I'm going to work on vehicle orientation based on track normal next, my track collision code is just checking min max right now, I removed my ray plane intersection code while debugging, I'll keep it simple until I get orientation working and I can go around the whole track

then I'll add gravitational torque back in and improve the controls

then I'll improve collisions and then I'll start making the track go upside down and have loops

[cloud rivet]

it might not look like it, but it is progress. I think I need to interpolate the normals across the tracks, and there's some orientational instability I need to resolve idk how yet

I might do a collision sphere

I probably should

[cloud rivet]

I like working on this project

[solid grove]

schrodinger's emotions

[cloud rivet]

I updated my iPhone to 26 with the glass ui . It is terrible

Swiping between views stutters

It’s like a 2010 era android experience

[solid grove]

what model of device?

[cloud rivet]

Maybe it’s obsolete

[solid grove]

26.1 is a lot better than 26.0 if you can believe it

they’ll probably fix it over time

[astral hinge]

26.7 is where it peaked

[solid grove]

time traveler spotted

[vagrant musk]

I’m still on 17.1.1

[broken fog]

huh ios 26 is fine for me on ipad

macos tahoe otoh is shiiiit

[vagrant musk]

I wish I was exactly 0.0.1 versions more up to date
no further

[broken fog]

18 was fine?

[vagrant musk]

Was 18 the siri update

[broken fog]

idk haven't used siri ever

[vagrant musk]

Because I know siri is like unusable in newer versions

[vagrant musk]

17.1.1 forces an annoying feature into the keyboard if I don’t turn off a different feature I do like
17.1.2 offers an option to turn it off independently

[cloud rivet]

Siri has always been unusable

[cloud rivet]

I upgraded because I bought airpod pro 3s and they are from the future. They are amazing. I had the pro 2s before

The 3s made me update ios to 26

[vagrant musk]

Ok but it at least doesn’t cancel the interaction 99% of the time in 17.1.1
Works pretty reliably for at least making calls and setting alarms back here

[broken fog]

they don't work with older ios?

[cloud rivet]

they were an impulse buy after I walked into an apple store, I came out of the store and sat on a bench to try to pair them and I got a warning saying it had features that required a software update to 26

[cloud rivet]

The only thing I have ever used siri for is using my apple watch to ask it to find my iphone and that's always worked

I'm not sure how to find my iphone using my watch without siri

if that doesn't work anymore lol, what is happening to apple

[broken fog]

a very apple move

that sucks

i mean tbf there would be no reason not to update to the latest os if it didn't kinda suck ass so it's not entirely unreasonable

but they should probably make the latest os suck less ass

though like i said ios 26 is perfectly fine for me on ipad, tahoe is the sus one

[cloud rivet]

I haven't updated my ipad yet

my phone is 3 years old so maybe that's just too old for updates

in the past I bought a device every other year

[bronze socket]

I've got an iphone se, like the original one with iphone 6 internals

ios 15.6

[cloud rivet]

based

that was a good phone iirc for the price

[bronze socket]

still works, been running it since I think 2017

not gonna get another apple product if I can help it, but damn do smartphones look more and more ass these days

[cloud rivet]

the new iphone air looks incredible

but apparently it is actually a bad phone

but it looks just amazing

it's like wafer thin

the camera and sound on it are what they had to sacrifice iirc and a bad camera on an iphone is just a just a massive wtf

I just don't know what's going on with Apple

[bronze socket]

its ok they will just replace the bad parts of your image with ai

[broken fog]

nah that should be fine

[cloud rivet]

apple like the one tech company that is not throwing AI into every thing

[broken fog]

iphones are good for 5-6 years of updates at least

[cloud rivet]

I was looking at vs 2026 and it was like mostly AI updates

[broken fog]

check battery health, they do massively slow down the phones via software when the battery degrades

[bronze socket]

might need to think about hodling some apple stock, wouldn't buy their products though

oh yeah my battery health is godawful but for a 7 year old phone what can you expect

8 year old*

[broken fog]

that kinda sucks imo, it just gets a huge camera bump and it means a flimsier phone with smaller battery

much rather have the chonky 16/17 pro

[vagrant musk]

they put AI into like 3 things and kinda did it lazily lol

[cloud rivet]

fuck AI features lol

[vagrant musk]

but at least it can be turned off

[broken fog]

i genuinely respect them for that, as much as i despise their blatant anti consumer practices

[bronze socket]

it speaks volumes about their internal strategic planning

[broken fog]

i can also respect that they're still primarily a computer company and not an ad company

[bronze socket]

it makes me oddly confident in their stock value surviving or even growing when the AI bubble contracts

[cloud rivet]

I don't know if anyone is going to survive this bubble

it's fucking huge

it's going to have a huge blast radius

[bronze socket]

do you know how to buy chinese treasury bonds, asking for a friend

[cloud rivet]

this bubble it's like 10x the size of the GDP of nations

I think you can probably buy into something that has chinese treasury bonds

probably like a large India corp

the market value of the AI companies is well over 10 trillion, nvidia alone is $5 trillion, it's probably in the 20 trillions

maybe this is how it all ends

it is ai that kills us, but not the actual ai, just the hype

[broken fog]

apple stock seems like a pretty safe bet

i don't think it's massively spiked in recent times, but i also don't see it crashing any time soon

though i would probably sell any stock if i'm getting the feeling the ai bubble is about to pop

[bronze socket]

yeah stock exposure feels kinda sketchy now

[broken fog]

invest in gold

[bronze socket]

I am, precious metals are my #1 holding outside of us tsy bonds

I am the kind of conservative investor you see in your nightmares

I wonder if there is enough publicly available info to calculate the AI exposure of a given company

[broken fog]

you could create a startup and use ai to do it

[cloud rivet]

so for my project's next steps, I think I'm back in a place to start making progress. I will fix the track detection for the purpose of orienting the vehicle, it's using just the normal of first track that fits into the bounds per ray and it is creating a weird orientation issue where it is clearly picking the wrong track. So I'll fix that, and then I will actually interpolate the the normals of two track quads so the points are not trying to add torque for different normals.

I think that will clean up the weird orientation we see in that video.

the next thing will be a collision fix, so the vehicle doesn't disappear into the track

with those things in place I think I'll be able to go around the full track without weird issues

I think it should also not try to apply the orientation torque while falling, I think that's why it's tumbling around at the beginning, it's trying to apply the torque based on the distance to the track below while closing the distance at a rapid pace and it ends up over correcting and then it takes a while to stabilize

[wraith urchin]

Have you been successfully radicalized against AI?

[cloud rivet]

I don't like AI, I have to use it at work, and I just vocally disagree with the statements statements that say it doesn't work, when I know it does work, and can work really well if used correctly. I am trying to correct what I see as statements that dismiss AI as bad when I have contrary evidence based on my experience.

when really skilled people use AI it is definitely super valuable, seeing people new to programming and graphics try to use it has the worst case outcomes because they are making a mistake to try and use AI for something they don't already now how to do so they don't know how to use AI and they don't know how to do graphics programming

[wraith urchin]

When the bubble pops they will scrap all the data centers and we won't be able to use it even if we wanted to

[cloud rivet]

I think if those of us here are basing our opinions about the effectiveness on AI based on how unskilled people are using it, you're going to draw incorrect conclusions

[cloud rivet]

one can hope

[wraith urchin]

Skilled people can just write good code though. No need for the AI

[cloud rivet]

they can, but you will save a ton of time if AI is used

for a subset of the things that you work on

that's all it is good for, doing things faster, and when it is faster, it is a lot faster

if you can do it faster yourself you should not use it

[solid grove]

There are lots of successful ML deployments every year in a wide variety of applications, but they receive little fanfare

[bronze socket]

imagine the secondary market of enterprise grade GPUs

that suddenly no one wants

[solid grove]

eg advancements in ML-based automatic CC for video are pretty awesome

and a great accessibility feature

[bronze socket]

yeah ML has plenty of uses but being your chatbot friend/slave isn't one of them

[solid grove]

the bubble is fueled entirely on companies betting that they can replace their workforces with AI

“you’re telling me i can have 0% expenses and 100% profit?!”

when the bubble pops all of the useful ML applications, like the CC thing, will stick around and continue to be made

[bronze socket]

there will hopefully be enough of a contraction to make GPUs a little cheaper and beefier for the home user though

[solid grove]

Apparently Microsoft bought more GPUs than they have power and data center space to run them

[bronze socket]

OG speculation and scalping strats, buy shit and stick it in a warehouse until it's worth something

hope they diversified into beanie babies

[solid grove]

unfortunately for them electronics don’t hold their value very well once the new one comes out

[wraith urchin]

I will be buying all of them

Although when the bubble pops autodesk will probably lose a ton of money and lay me off

So maybe I won't have the excess cash for that...

[cloud rivet]

we're hiring

[solid grove]

will u be when the bubble pops?

[cloud rivet]

yes

we raised so much cash

we'll outlive the bubble

[wraith urchin]

It's OK im looking forward to being layed off, more time for ASO

[cloud rivet]

people's obligation to audit access wont' go away with a bubble, and we save customers a ton of money and time