@olafinho.xyz
π€ 56
π₯ 658
π 298
pinned post!
How to install multiple CUDA versions on single machine and use them conveniently in different envs? -> (this might be useful when installing many codebases locally, (without some form of isolation), each with different torch+cuda req.):
7 months ago
1
0
0
writing notes on paper is great, until you forget to take notepad from home and are in dire need informations from previous days π
10 days ago
0
1
0
11 days ago
0
2
0
reposted by
Nathan Lambert
12 days ago
Use multiple models Why I'm using all of GPT 5.2 Thinking/Pro, Claude Opus 4.5, Gemini 3, and occasionally even Grok. How my usage has changed over time. The meta for getting the most out of AI in 2026.
www.interconnects.ai/p/use-multip...
loading . . .
Use multiple models
The new meta for getting the most out of AI in 2026.
https://www.interconnects.ai/p/use-multiple-models
2
33
4
robot folding laundry; claim: 200,000 garments from ten different clients in the past 3 months
www.dyna.co/blog/monster...
loading . . .
DYNA - Production-Ready Foundation Model Robot
DYNA: The first production-ready foundation model robot β built to reliably deploy in the wild.
https://www.dyna.co/blog/monster-laundry
14 days ago
0
1
0
i like having thoughts
25 days ago
0
1
0
Lion does not concern himself with "Your disk is almost full. Save space by optimising storage."
about 1 month ago
0
2
0
This seems to be extremely cool! will have to find time to set it up
meshsplatting.github.io
about 1 month ago
0
3
2
interesting what happened to cloudflare reliability after this post π€
add a skeleton here at some point
about 2 months ago
0
0
0
the fact i do not remember material i learned month ago is the real catastrophic forgetting
about 2 months ago
0
2
0
reposted by
Breaking Taps
about 2 months ago
The last month of my life: BreakingTTAPs This is a custom transport-triggered, 32 bit processor that will be fabricated by GlobalFoundries on their 180nm process (via of
wafer.space
) I'll make a video at some point, but some high level details here:
2
31
7
important question is being asked on the other site
about 2 months ago
0
3
0
listening rn
www.youtube.com/watch?v=aR20...
loading . . .
Ilya Sutskever β We're moving from the age of scaling to the age of research
YouTube video by Dwarkesh Patel
https://www.youtube.com/watch?v=aR20FWCCjAs
about 2 months ago
0
2
0
handcrafting gcode has never been easier than rn due to llms and i think it's awesome
about 2 months ago
0
0
0
what is software equivalent to wago connectors
2 months ago
0
0
0
one must imagine sisyphus happy
2 months ago
0
0
0
cloudflare pages is cool i guess
2 months ago
0
0
1
youtu.be/lVvURVG_b10?...
loading . . .
NG-2 Launch Recap
YouTube video by Blue Origin
https://youtu.be/lVvURVG_b10?si=4jZZrK8gt4jp6RV6
2 months ago
0
1
0
marble from WorldLabs is officially available to all!
marble.worldlabs.ai
loading . . .
Marble | World Labs
Create and share 3D worlds with Marble
https://marble.worldlabs.ai/
2 months ago
0
1
0
please get me out of MLOps mines.......
2 months ago
0
0
0
oh wow, k-scale labs is closing:(
3 months ago
1
1
0
me looking at the soup of splats generated by my broken code
3 months ago
0
1
0
dear diary, the cheap glue dispenser controller arrived; after first test i can say it's promising
add a skeleton here at some point
3 months ago
0
0
0
reading something multiple times until understanding is underrated
3 months ago
0
0
0
putting my codebase into dishwasher to make it clean code
3 months ago
0
0
0
Simclusters algo has a potential
3 months ago
0
0
0
all images should be allowed to only exist in 1:1 RATIO, S Q U A R E S also: exif rotation header? be serious
3 months ago
0
0
0
interesting discussion thread on colmap vs vggt,etc on the other site:
x.com/gabriberton/...
loading . . .
Gabriele Berton on X: "Is COLMAP still widely used or are Mast3r / VGGT taking over? https://t.co/For5yIzDXq" / X
Is COLMAP still widely used or are Mast3r / VGGT taking over? https://t.co/For5yIzDXq
https://x.com/gabriberton/status/1983764005027262760
3 months ago
0
0
0
that feeling when preliminary results look promising, so you hesitate to move forward- to not shatter the illusion of hopium
3 months ago
1
0
0
reposted by
Tim Kellogg
3 months ago
Marin β an open lab for developing AI in public - they just released Marin 32B, a competitive base model - all experiments & results are public - heavily funded by Google & TPUs
marin.community
1
32
3
reposted by
Grace
9 months ago
Remember when we all thought Ilya Sutskever had seen the machine god in the depths of openai? Good times
7
67
5
please For you algo, have mercy on me and show this to some CV folks: what is the best oss target based camera calibration toolbox/project? mc-calib?
3 months ago
0
0
0
yt music > apple music/spotify, dont @ me, it's simply true
3 months ago
0
0
0
dear diary: i've decided, after automatic glue station my next sideproject is gonna be buliding AmazingHand from huggingface/pollen
3 months ago
0
0
1
so, about that. i called a company that deals with distribution/selling and occasional workshops about adhesives and talked with the owner. damn it is delightful to listen someone competent passionately talk about their domain's subject
add a skeleton here at some point
3 months ago
1
0
0
imagine hating someone's ideas so much you make a diss painting Blake's opposition to the Enlightenment was deeply rooted., Blake wrote "Art is the Tree of Life. Science is the Tree of Death." Newton, whose "natural religion" of scientific materialism he characterized as sterile." (quotes cut)
3 months ago
1
2
0
alright, everyone STOP. let's settle this: what is the definition of world model because i am going to lose my mind
3 months ago
1
1
0
Babe wake up, Breaking Taps dropped new vid
youtu.be/Y3h0vZ2PYl8?...
loading . . .
Making a (Robotic) Muscle is Hard
YouTube video by Breaking Taps
https://youtu.be/Y3h0vZ2PYl8?si=eQn-mjC6lN44FKmF
3 months ago
0
0
0
gm, it's another beautiful day to work on doohickeys
4 months ago
0
0
0
does anyone use arxiv html viewer?
4 months ago
0
1
0
ok chat, this paper seems to be interesting:
shikun.io/projects/kal...
loading . . .
Kaleido
Scaling Sequence-to-Sequence Generative Neural Rendering
https://shikun.io/projects/kaleido
4 months ago
0
0
0
Iβm ashamed to admit 5 months ago I bought lipo battery charger for hobby projects and still havenβt used it to this day
4 months ago
0
0
0
crazy to think that people used to bind books with human skin (sometimes as a form of a bonus punishment to convicted people), human-covered book, wth
4 months ago
1
2
0
sometim number go up; sometim go down
4 months ago
0
0
0
another beautiful day to learn something, hell yeah
4 months ago
0
0
0
it's over. (coke zero run out)
4 months ago
0
0
0
i learn, then i forget, i learn then i forget, i learn, then i forget, i learn then i forget, i
4 months ago
0
0
0
since this post was a viral success, another one: any glue expert here??
add a skeleton here at some point
4 months ago
0
0
1
so, uhh, does anyone here know a thing about budget cnc plasma cutting or is this the wrong room?
4 months ago
0
0
1
reposted by
David Picard
4 months ago
I coined a new verb on my slack channel. Suttonize: verb. Removing dedicated processing that require careful tuning and replacing it with simpler but more intensive computation. Ex: "We should suttonize this: just remove the encoder and concat the sources together with a few self-attention layers".
4
35
7
"All it takes is for the rendered image to look right" - Jim Blinn
4 months ago
0
0
0
Load more
feeds!
log in