kache

@yacineMTB • 411,310 subscribers

RL, robots, embedded, simulators. prev eng @ x, stripe. 6'3 (height) tensorpunk, Father 2u2 subscribe to read my blog!

Shorts

180,624 views

55,794 views

154,861 views

68,080 views

61,802 views

84,475 views

94,391 views

29,165 views

68,441 views

33,282 views

39,217 views

45,350 views

136,085 views

15,989 views

138,398 views

35,603 views

54,367 views

31,586 views

30,204 views

Videos

sweetdream.ai

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Private Show

Join now for exclusive access

Free preview available • Premium content

789,051 views • 2 months ago

172,995 views • 20 days ago

745,260 views • 3 months ago

834,182 views • 4 months ago

557,924 views • 4 months ago

1,208,755 views • 1 year ago

241,670 views • 2 months ago

129,632 views • 1 month ago

1,652,675 views • 2 years ago

98,733 views • 1 month ago

114,777 views • 2 months ago

102,116 views • 1 month ago

65,272 views • 1 month ago

133,150 views • 3 months ago

376,887 views • 10 months ago

128,536 views • 3 months ago

147,323 views • 4 months ago

45,392 views • 1 month ago

31,749 views • 1 month ago

91,279 views • 3 months ago

Live Cam

kache

Shorts

this component is insane i have been waiting for it for 2 years

I am streaming pictures to the drone from a random drone video game (liftoff). Over the air to this tiny board Using only the pictures, the AI is able to fly in the game, in Acro mode

Dennis Gustafsson me and gpt 5.5 found a bug that didn't make it deterministic and fixed it

pretty amazing what you can do in a web browser if you know what you're doing

I just gave up and changed my personal website into a forum that AI agents can use to collude. Might as well just get it over with

neural networks are truly amazing

I wrote my own CAD software whose intended user is an AI. I asked GPT 5.6 sol to make me a case for my camera. It only interrupted me once, which was to log in to a website to download a model of my camera as reference. Not perfect but.. pretty crazy :)

i am having way too much fun

I just trained cartpole in mujoco at 18 million steps per second. This policy learned in **less than 3 seconds** rollout policy batch size was 8192 agents

i kind of can't believe it

"how many experiments have you run" yes

configuring nvim for 5 hours today was worth it look at this cute lil guy

i got annoyed with bambulabs so i replaced it now i don't need bambulabs anymore

i want to show you guys how fast programming is now i'm kind of resurrecting dingboard zig to see how fast i can make progress on it, using it as a test bed to see AI progress it took me 5 minutes to add a GPU accelerated (!) mirror button now, watch how fast this next thing is

researcher friend of mine shared this with me; comfree-sim. which is basically existence proof of my hunch: mujoco has terrible performance, and there are probably huge huge gains you can make in training, especially considering domain randomization

this is a legal move in the software that they use you can define it to be 3v3 -&gt; cap -&gt; cap -&gt; MCU in the schematic but it will still allow you to do this

what if you simulated every single particle of your robot would it be parallelizable then?

I've gotten a mujoco sim RL training loop for a unitree robot at 200k SPS. I'm looking into the physics for friction, contact dynamics. My goal: can I reproduce &amp; beat the mujoco playground RL baselines This is running in my web browser with raylib. Its the baseline

Videos

Watch Anya Live

Melee.c - super smash bros melee, decompiled and reverse engineered into 5000 lines or so of c All done with deepseek flash 0731, from my phone, while I'm hanging out with my newborn

i have a fully automated pipeline that turns real places into 3d environments. I've been running it for weeks now. It might not be obvious, but all of these are assets specialized. They aren't meant to be rendered normally

Fable and I rewrote Box3d entirely in native cuda. It's about 30x faster on my GPU compared to my CPU (depending on the situation). This was my hardest benchmark. I still had to be there, but soon I won't. Cost me about 200 dollars

When I worked at twitter; I got ping'd by someone from the bot detection team asking me if my usage was natural -- I hit one of their new test bot detectors. Then I sent him a video of how I use this website and he was like "oh.."

behold. THE WORLDS FIRST SIX PENDULUM CARTPOLE SOLVE. Including a sponsor! To solve this task, I built an environment to train an AI. This is what mechanize does, but for larger AIs. Apply! Salaries are up on their page Thank you to mechanize for sponsoring!

everyone is building humanoid robots when they should be building this

all done with nothing but deepseek flash 0731 reverse engineered the animation system built my own renderer on top of it set up debug views so i can understand what makes it so magical

Guess whose state of the art rigid body simulator that can support up to 200k rigidbodies is accurate enough for the microduck to walk in Not a single bit of this is run with mujoco. This is entirely my rigid body simulator (based on box3d) and runs entirely on cuda

programming in 2024

new high score. average 27.8 FPS depth all from an orange pi zero. a board the side of three quarters i actually can't believe i've been able to push it this far. this is absurd

it's honestly incredible how bad these things are at programming. like they do the worst thing possible 100% of the time, no matter how "smart" they get. so i have to always read the code. and always correct them. its so tiresome. its getting better at least..

the singularity is here

I'm launching Dingcad Just copy the prompt and put it into codex. It's free. Have fun!

after having sol rip on it all night, i can do over twice as many boxes without dropping to unusable fps

wrote a quick script to scan through the accounts i've blocked and which country they're posting from

this is GPU accelerated, no circular dependency physics "solver" (there is nothing being solved) each particle gets a cuda thread (not literally but you get what i mean). Then it gets reduced up to what it should resolve to regarding the rigid bodies

my website is now live editable by me. While I edit it, you can see the edits happen live

this entire thing is done on an orange pi zero 3w. no more 4090. stereo depth inference at sub 60ms by custom writing NPU kernels for it 14 gram board, 30mm x 65mm pcb

All assembled and streaming a depth map to my phone from the basement to my bedroom I am going to make 4 of them. Anyone wanna buy?

scales great! each rectangle here is a set of discs. running on cuda. now to painfully go through all of the weird collision bugs and figure out why they're happening and why it isn't settling

I just trained cartpole in mujoco at 18 million steps per second. This policy learned in less than 3 seconds rollout policy batch size was 8192 agents

this is a legal move in the software that they use you can define it to be 3v3 -> cap -> cap -> MCU in the schematic but it will still allow you to do this

I've gotten a mujoco sim RL training loop for a unitree robot at 200k SPS. I'm looking into the physics for friction, contact dynamics. My goal: can I reproduce & beat the mujoco playground RL baselines This is running in my web browser with raylib. Its the baseline