Загрузка видео...

Не удалось загрузить видео

На главную

I've been wanting proper Gaussian Splatting support in Rerun for the longest time. I had previously vibe-coded a custom view that kind of worked before, but it had some wonky bugs and was still too rough around the edges. This was originally done back in the Codex 5.4 era,...

pablovelagomez1's profile picture

Pablo Vela

2,889 subscribers

18,842 просмотров • 1 месяц назад •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

There's been a few cool updates recently. In particular, Rerun 0.33 released headless rendering. This, along with the Fable 5 release pushed me to work torwards making MAMMA realtime! I threw Fable at the problem, and it was able to take original implementation that was ~12 seconds / frame and get it all the way down to 40ms /frame, or nearly a 300x speedup 🏎️ How did I achieve this? TLDR: - Use rerun's headless rendering as supervision when optimizing - Save rrd file as test fixture to guide model optiziation with /goal - create an html artifact with headless rendering to provide detailed breakdown of what it did and how it actually looks like in the viewer There were a few critical bits to make sure that this ACTUALLY worked and that Fable didn't just cheat or delete something and declare victory. The first is that the original version used Rerun, this allowed us to save things to disk as an RRD file, meaning we could query the contents and use this as a sort of test fixture or golden artifact that held EXACTLY what all of the values should be. Then we can use this with /goal as a metric when doing the optimization to ensure there are no regressions. The second bit is the headless rendering, this gave us the ability to check that not only did the test fixture pass, but it also looked visually correct. This made a huge difference, and an awesome side affect of it is that we can use the headless rendering to create an implementations.html file. This gives a visual guide as to what the agent did (I walk through it in the video below) Along with this, we're working on an MCP server for rerun that allows full interactivity with the rerun viewer for your agent. So for example the agent can click, drag, move views, scroll timelines, ect. I used this to help the agent debug certain parts such as when the 2d sam masks didn't line up, or if the triangulated keypoints werent correctly matching with the optimized mesh. The agents could go, click into the view, scroll through the timeline and see where things went wrong. Fable + Headless Rendering + Rerun MCP == 300x speedup in less then a days work With these new tools, I'm planning on going back to my gaussian splatting implemntation and cleaning it up + making it fast!
1:33

There's been a few cool updates recently. In particular, Rerun 0.33 released headless rendering. This, along with the Fable 5 release pushed me to work torwards making MAMMA realtime! I threw Fable at the problem, and it was able to take original implementation that was ~12 seconds / frame and get it all the way down to 40ms /frame, or nearly a 300x speedup 🏎️ How did I achieve this? TLDR: - Use rerun's headless rendering as supervision when optimizing - Save rrd file as test fixture to guide model optiziation with /goal - create an html artifact with headless rendering to provide detailed breakdown of what it did and how it actually looks like in the viewer There were a few critical bits to make sure that this ACTUALLY worked and that Fable didn't just cheat or delete something and declare victory. The first is that the original version used Rerun, this allowed us to save things to disk as an RRD file, meaning we could query the contents and use this as a sort of test fixture or golden artifact that held EXACTLY what all of the values should be. Then we can use this with /goal as a metric when doing the optimization to ensure there are no regressions. The second bit is the headless rendering, this gave us the ability to check that not only did the test fixture pass, but it also looked visually correct. This made a huge difference, and an awesome side affect of it is that we can use the headless rendering to create an implementations.html file. This gives a visual guide as to what the agent did (I walk through it in the video below) Along with this, we're working on an MCP server for rerun that allows full interactivity with the rerun viewer for your agent. So for example the agent can click, drag, move views, scroll timelines, ect. I used this to help the agent debug certain parts such as when the 2d sam masks didn't line up, or if the triangulated keypoints werent correctly matching with the optimized mesh. The agents could go, click into the view, scroll through the timeline and see where things went wrong. Fable + Headless Rendering + Rerun MCP == 300x speedup in less then a days work With these new tools, I'm planning on going back to my gaussian splatting implemntation and cleaning it up + making it fast!

Pablo Vela

22,880 просмотров • 2 месяцев назад

Colmap 4.0 was very recently released, so it inspired me to do some work to better understand it and its new capabilities with Rerun. I want to really understand how Colmap, and in particular, pycolmap, works outside of just calling it via the CLI. So my goal is to use the low-level pycolmap API to log every part of the pipeline. The explicit goal is to have an alternative to the SQLite database that I can utilize. Instead of SQLite, I want to try logging everything directly to rerun and use RRD. This means I can have deep inspectability and still save the features/matches/2D view geometry, but be able to view it directly in rerun. I think this is one of the superpowers that rerun provides; data and visualizations are deeply integrated. As I'm often working with sequential data (videos), I'm going to specifically focus on four things: 1. Monocular Video Simple: Calls high-level APIs such as pycolmap.extract_features, pycolmap.match_sequential, pycolmap.incremental_mapping. These are basically identical to the CLI options and provide a good baseline. 2. Monocular Video Streamed: Take the above high-level APIs and break them down to their iterator version, logging each component in a streamed manner. This way, I can stream the intermediate features to rerun while the extraction/matching/mapping is happening. 3. Rig with unknown calibration: <- WHAT THE VIDEO SHOWS This is probably the most interesting version and the first one I've been working on. It allows one to set a rig between known sensors, such as in VR/AR devices, leading to much better reconstructions with multiple cameras. This is the case where we don't know the calibration a priori, so we have to run a reconstruction twice: once as a normal Colmap reconstruction with no rig constraints, use this to generate the constraints, and then do it again with the newly found rig. 4. Rig with known calibration: This is the RoboCap example, where we have a pre-calibrated set of sensors, so we don't need to run the two reconstructions and also gain better matching between cameras, both spatially and temporally. Again, this leads to a much better reconstruction! Along with all this, GLOMAP has become a first-class global mapper, making it super easy to use directly within pycolmap! I'm excited to do more with this and compare it to things like pycuvslam, vipe, and other alternatives.
0:20

Colmap 4.0 was very recently released, so it inspired me to do some work to better understand it and its new capabilities with Rerun. I want to really understand how Colmap, and in particular, pycolmap, works outside of just calling it via the CLI. So my goal is to use the low-level pycolmap API to log every part of the pipeline. The explicit goal is to have an alternative to the SQLite database that I can utilize. Instead of SQLite, I want to try logging everything directly to rerun and use RRD. This means I can have deep inspectability and still save the features/matches/2D view geometry, but be able to view it directly in rerun. I think this is one of the superpowers that rerun provides; data and visualizations are deeply integrated. As I'm often working with sequential data (videos), I'm going to specifically focus on four things: 1. Monocular Video Simple: Calls high-level APIs such as pycolmap.extract_features, pycolmap.match_sequential, pycolmap.incremental_mapping. These are basically identical to the CLI options and provide a good baseline. 2. Monocular Video Streamed: Take the above high-level APIs and break them down to their iterator version, logging each component in a streamed manner. This way, I can stream the intermediate features to rerun while the extraction/matching/mapping is happening. 3. Rig with unknown calibration: <- WHAT THE VIDEO SHOWS This is probably the most interesting version and the first one I've been working on. It allows one to set a rig between known sensors, such as in VR/AR devices, leading to much better reconstructions with multiple cameras. This is the case where we don't know the calibration a priori, so we have to run a reconstruction twice: once as a normal Colmap reconstruction with no rig constraints, use this to generate the constraints, and then do it again with the newly found rig. 4. Rig with known calibration: This is the RoboCap example, where we have a pre-calibrated set of sensors, so we don't need to run the two reconstructions and also gain better matching between cameras, both spatially and temporally. Again, this leads to a much better reconstruction! Along with all this, GLOMAP has become a first-class global mapper, making it super easy to use directly within pycolmap! I'm excited to do more with this and compare it to things like pycuvslam, vipe, and other alternatives.

Pablo Vela

30,070 просмотров • 4 месяцев назад

This soldiering training is the most impressive and immersive I've ever tried. It is 10x better than a YouTube tutorial video. It really allowed me to see the procedure from all the points of view, and even get super close to get the details. It is a collaboration between Gracia and Imperial College London: they recorded a soldering session with Gaussian Splatting Videos (4DGS), so that you can enjoy it from your VR headset. You can see the action happening in front of you; you can pause and re-watch what you need, change the point of view, get closer, get more distant. And the quality with which it has been shot is impressive: I enjoyed this experience with my DELL Pro Max Tower T2 with NVIDIA Pro RTX 6000, and I could really see all the small details of the PCB that was being soldered! I was really impressed by it. But I also noticed some drawbacks. First of all, some scenes have artifacts that make seeing the details of the PCB hard. I think when it comes to training involving small details, the capture and reproduction of the splat should be flawless. Then, as much as I loved it as a passive experience, I would have liked to have also some sort of practice session in VR. The power of VR is to let you learn by doing in full safety, and this kind of training does not exploit it. But maybe the best way to enjoy it would be in MR, where you have this training video close to a real workbench where you do the actual soldering while following the tutorial. Gaussian Splat can really revolutionize training: this kind of recording is much better than any flatscreen experience, and more accurate than any 3D CGI recostruction. I suggest you give it a go at this experience in the Gracia app (it's free). Then let me know your impressions! #VirtualReality #training #DellProPrecision #GaussianSplat #technology [Disclaimer: I'm a DELL Pro Precision Ambassador, and this is why I mentioned the model of my PC. I have been given a PC to do cool tests and share my results on social media. No monetary compensation or sales affiliation is part of the collaboration]
0:31

This soldiering training is the most impressive and immersive I've ever tried. It is 10x better than a YouTube tutorial video. It really allowed me to see the procedure from all the points of view, and even get super close to get the details. It is a collaboration between Gracia and Imperial College London: they recorded a soldering session with Gaussian Splatting Videos (4DGS), so that you can enjoy it from your VR headset. You can see the action happening in front of you; you can pause and re-watch what you need, change the point of view, get closer, get more distant. And the quality with which it has been shot is impressive: I enjoyed this experience with my DELL Pro Max Tower T2 with NVIDIA Pro RTX 6000, and I could really see all the small details of the PCB that was being soldered! I was really impressed by it. But I also noticed some drawbacks. First of all, some scenes have artifacts that make seeing the details of the PCB hard. I think when it comes to training involving small details, the capture and reproduction of the splat should be flawless. Then, as much as I loved it as a passive experience, I would have liked to have also some sort of practice session in VR. The power of VR is to let you learn by doing in full safety, and this kind of training does not exploit it. But maybe the best way to enjoy it would be in MR, where you have this training video close to a real workbench where you do the actual soldering while following the tutorial. Gaussian Splat can really revolutionize training: this kind of recording is much better than any flatscreen experience, and more accurate than any 3D CGI recostruction. I suggest you give it a go at this experience in the Gracia app (it's free). Then let me know your impressions! #VirtualReality #training #DellProPrecision #GaussianSplat #technology [Disclaimer: I'm a DELL Pro Precision Ambassador, and this is why I mentioned the model of my PC. I have been given a PC to do cool tests and share my results on social media. No monetary compensation or sales affiliation is part of the collaboration]

TonyVT SkarredGhost

37,876 просмотров • 11 дней назад

We have HOT3D! I've started using Claude to port more datasets into Rerun and exoego-forge. I'd been meaning to bring in the HOT3D dataset from Meta for a while, but with Claude, it's way easier. My goal is to take any egocentric, exocentric, or both datasets and ingest them into a standardized schema. Getting everything into Rerun means we can easily query and transform data via the in-memory OSS server. This lets us generate SQL-like queries such as: "Find me all frames that only contain left hands in the leftmost camera view." Most people think of Rerun as a viewer, but this is the actual superpower. So far we have: 1. HOT3D 2. Hocap 3. UmeTrack 4. Assembly101 5. EgoDex Planning to add more, and with every addition, it gets easier as we build up agent skills and better code examples. Hoping to make it almost fully automatic for adding new datasets. The next few I'm looking at are Harmony4D and Aria Pilot Gen2 After we have enough samples, I'll work on bringing in all the different algorithms I've worked on to transform the data 🙂
0:29

We have HOT3D! I've started using Claude to port more datasets into Rerun and exoego-forge. I'd been meaning to bring in the HOT3D dataset from Meta for a while, but with Claude, it's way easier. My goal is to take any egocentric, exocentric, or both datasets and ingest them into a standardized schema. Getting everything into Rerun means we can easily query and transform data via the in-memory OSS server. This lets us generate SQL-like queries such as: "Find me all frames that only contain left hands in the leftmost camera view." Most people think of Rerun as a viewer, but this is the actual superpower. So far we have: 1. HOT3D 2. Hocap 3. UmeTrack 4. Assembly101 5. EgoDex Planning to add more, and with every addition, it gets easier as we build up agent skills and better code examples. Hoping to make it almost fully automatic for adding new datasets. The next few I'm looking at are Harmony4D and Aria Pilot Gen2 After we have enough samples, I'll work on bringing in all the different algorithms I've worked on to transform the data 🙂

Pablo Vela

35,662 просмотров • 4 месяцев назад

"why? just hard work from my engineers and from myself to improve me as a driver" 👏🏻👏🏻 🗣️: friday night did you think you had the car to win the race? and do you have an explanation of the difference of pace between you and oscar? lando: "no, i didn't think so. especially after fp1, i didn't. actually, andrea wanted to make a bet with me because i was pretty disappointed with the pace and how the car was handling in fp1. it got a bit better in fp2, but still the ferraris were just on another level. and i really didn't expect to get as close as we did in quali. but the wind changed from friday to saturday, and the saturday wind really suited our car a lot more. we had a lot more confidence in driving the car, so i could push more. and then i certainly felt like i could be on pole yesterday, and i felt like i could win a race today. i think today the race pace was much better than i was expecting. the degradation was much better than i was expecting too. but it was probably one of my best performances personally in terms of driving the car. the fact i can keep up with oscar on this kind of track, where it's impossible to follow, impossible to overtake, i was within eight tenths or something for 40 laps almost. the fact i could do this is what gave me confidence that i would win the race no matter what, whether he boxed early or late or whatever. the pace advantage i had was huge today. but i don't know why. the cars are identical. and i've just been feeling very, very strong lately. i've worked on many areas from last year. i feel much more confident lately with how to extract performance. and i worked a lot with my engineers to understand this car, because it's certainly not an easy car to understand and figure out on how to drive on the limit consistently. so why? just hard work from my engineers and from myself to improve me as a driver and improve my driving. and that is the answer"
2:17

"why? just hard work from my engineers and from myself to improve me as a driver" 👏🏻👏🏻 🗣️: friday night did you think you had the car to win the race? and do you have an explanation of the difference of pace between you and oscar? lando: "no, i didn't think so. especially after fp1, i didn't. actually, andrea wanted to make a bet with me because i was pretty disappointed with the pace and how the car was handling in fp1. it got a bit better in fp2, but still the ferraris were just on another level. and i really didn't expect to get as close as we did in quali. but the wind changed from friday to saturday, and the saturday wind really suited our car a lot more. we had a lot more confidence in driving the car, so i could push more. and then i certainly felt like i could be on pole yesterday, and i felt like i could win a race today. i think today the race pace was much better than i was expecting. the degradation was much better than i was expecting too. but it was probably one of my best performances personally in terms of driving the car. the fact i can keep up with oscar on this kind of track, where it's impossible to follow, impossible to overtake, i was within eight tenths or something for 40 laps almost. the fact i could do this is what gave me confidence that i would win the race no matter what, whether he boxed early or late or whatever. the pace advantage i had was huge today. but i don't know why. the cars are identical. and i've just been feeling very, very strong lately. i've worked on many areas from last year. i feel much more confident lately with how to extract performance. and i worked a lot with my engineers to understand this car, because it's certainly not an easy car to understand and figure out on how to drive on the limit consistently. so why? just hard work from my engineers and from myself to improve me as a driver and improve my driving. and that is the answer"

ray

55,185 просмотров • 27 дней назад

Q: It must be complicated, when I listen to you, to have a private life, somebody to understand your passion and to share this moment. Lewis: "It really is, especially I would say more so today than ever before, which is the way the world is, you know. I look at the other drivers and I wonder how they're doing it. You know, some are having kids and some married, some, you know, most of them girlfriends. I did that when I was in my 20s, but I took a decision to really to maximize my time that I have here because it's not as long as you think and it's limited, you know. And I don't want to look back and be like, ah, if I just gave a little bit more here, I didn't sacrifice my time because I was committed elsewhere." "So I really focused in these last, you know, particularly these last 10 years, like get everything I can out of my performance. Then when I retire, then I can do whatever I want. You know, I can dedicate my time to whatever else it is and not have to worry." "But in this competition time, focus on health, well-being, my mental health, my driving technique, being as good an engineer as I can be, and also being the best teammate that I can potentially be for the guys that I get to work with. That's my sole focus. You know, I want to win." "I've been fortunate enough to win with great teams in the past. Particularly, obviously, with Mercedes and with McLaren, which was incredible. And my dream is to win a championship with Ferrari." "And that's something that hasn't been done for a while. But they have absolutely every ingredient that's needed to win. It's just like getting all the pieces of the puzzle in the right place. And that's what I'm trying to work on in the background with Fred and the whole team." [📹 VIGNERON GAETAN]
2:02

Q: It must be complicated, when I listen to you, to have a private life, somebody to understand your passion and to share this moment. Lewis: "It really is, especially I would say more so today than ever before, which is the way the world is, you know. I look at the other drivers and I wonder how they're doing it. You know, some are having kids and some married, some, you know, most of them girlfriends. I did that when I was in my 20s, but I took a decision to really to maximize my time that I have here because it's not as long as you think and it's limited, you know. And I don't want to look back and be like, ah, if I just gave a little bit more here, I didn't sacrifice my time because I was committed elsewhere." "So I really focused in these last, you know, particularly these last 10 years, like get everything I can out of my performance. Then when I retire, then I can do whatever I want. You know, I can dedicate my time to whatever else it is and not have to worry." "But in this competition time, focus on health, well-being, my mental health, my driving technique, being as good an engineer as I can be, and also being the best teammate that I can potentially be for the guys that I get to work with. That's my sole focus. You know, I want to win." "I've been fortunate enough to win with great teams in the past. Particularly, obviously, with Mercedes and with McLaren, which was incredible. And my dream is to win a championship with Ferrari." "And that's something that hasn't been done for a while. But they have absolutely every ingredient that's needed to win. It's just like getting all the pieces of the puzzle in the right place. And that's what I'm trying to work on in the background with Fred and the whole team." [📹 VIGNERON GAETAN]

sim

86,907 просмотров • 1 год назад

When I saw the mask "Tribes of the Calf" from Kanbas I knew I had to make it into reality. The jewelry and gold really made it stand out for me. Since Sam Spratt's The Masquerade was revealed, I have been spending time sculpting and dissecting the mask to recreate it in 3D as faithfully as possible. I delved into the creation of this mask for many reasons. I love a good challenge and this mask surely was one for me. Creating something in 3D from a 2D image is not easy, and especially when the source has generative nature, some stuff is hard to interpret, but I tried my best to make sure the visual integrity of the mask is as close to the original as possible. Splitting the whole mask into parts, filling the missing pieces so I can build the textures was quite a lot of work. I tried to present the mask in my own style with a slightly different colorway to adapt to the mask itself. Please enjoy this short animation, and turn on sound🔊 This piece is my statement that I am here to stay. That I have a voice that often feels being lost in the void. That I have been creating and posting digital art for over 20 years now and will continue until I'm gone. I have a story to tell and I want to be heard. The space we have here is small, and is shrinking day by day. It doesn't have to be like that. We need to support each other and push ourselves and people here, otherwise we are all doomed. As Kanbas has put in their observation of the mask: "Inspirational. Emotional. Natural." This is what our space can be, and this is me making a statement with this homage. I will share a 4k still below as well as a short video showing the 3D GLB interactive model together with a yt link to the 4k video since compression here is pretty bad.
0:40

When I saw the mask "Tribes of the Calf" from Kanbas I knew I had to make it into reality. The jewelry and gold really made it stand out for me. Since Sam Spratt's The Masquerade was revealed, I have been spending time sculpting and dissecting the mask to recreate it in 3D as faithfully as possible. I delved into the creation of this mask for many reasons. I love a good challenge and this mask surely was one for me. Creating something in 3D from a 2D image is not easy, and especially when the source has generative nature, some stuff is hard to interpret, but I tried my best to make sure the visual integrity of the mask is as close to the original as possible. Splitting the whole mask into parts, filling the missing pieces so I can build the textures was quite a lot of work. I tried to present the mask in my own style with a slightly different colorway to adapt to the mask itself. Please enjoy this short animation, and turn on sound🔊 This piece is my statement that I am here to stay. That I have a voice that often feels being lost in the void. That I have been creating and posting digital art for over 20 years now and will continue until I'm gone. I have a story to tell and I want to be heard. The space we have here is small, and is shrinking day by day. It doesn't have to be like that. We need to support each other and push ourselves and people here, otherwise we are all doomed. As Kanbas has put in their observation of the mask: "Inspirational. Emotional. Natural." This is what our space can be, and this is me making a statement with this homage. I will share a 4k still below as well as a short video showing the 3D GLB interactive model together with a yt link to the 4k video since compression here is pretty bad.

shoneec

17,888 просмотров • 1 год назад

EMIBONNIE MOONSHADOW BLOOM #BBFanFest2026D2 🦊: First of all, I have to thank Bonnie. As you said, I'm not actually confident in myself at all, but it's mostly bc of you, who came and filled in the parts of me that I couldn't do at all, who came and filled in the parts that I didn't want to do at all in the beginning bc I was too afraid. It's bc of you that I am who I am today. If it weren't you, then it couldn't be anyone else. Thank you so much for always helping me in everything. For our show today, even though it's our second time, at first I was like, "It's going to be chill. Of course we can do it. We could do a full hour concert with just the two of us, this will pass in no time, we can do it!" But it turned out harder than every other time for many reasons. We had many things to do too. But I was willing to do everything and I was honestly very happy during the show. I feel like I don't usually get to see myself let myself go, let loose, and just chill to this extent. I really enjoyed this time and that I was brave enough to try many new things. 🦊: Thank you, Bonnie, and thank you to all my friends here for having fun together. This time, before going on stage, the feeling was like, "Alright! Let's do it, let's get on stage!" It felt really chill. Usually, I put a lot of pressure on myself. I'm glad that this time, I have a lot of friends here to help fill in the things that in the past I never thought I would've been able to do. Thank you so much, everyone. Thank you for having fun and letting loose together. 🦊: I'd also like to thank the executives for always giving me opportunities. Like what others have said, I don't feel that confident at all, but when I get opportunities and the executives believe that I can do it, it really makes me feel more confident. Thank you so much for giving me many opportunities in many different things and for giving them to me since the start. Truly, thank you so much. 🦊: Thank you as well to the crew. Actually, things behind the scenes are very hectic. There are a lot of things that have to be done this time, and they've all been very occupied taking care of us. Thank you for taking care of us and supporting us, both the staff in front of and behind the scenes, including the control technicians. 🦊: Lastly, thank you to all the fans. Without you all since day one, we wouldn't be here today as the Blush Blossom girls. Truly, thank you so much for supporting us even if you didn't know what this was going to be. Ever since the first time you've supported us, we've grown and expanded and gotten this big now. Thank you so much. 🦊: And thank you to myself as well. I'm so glad that I was brave enough to do this. Thank you.
4:42

EMIBONNIE MOONSHADOW BLOOM #BBFanFest2026D2 🦊: First of all, I have to thank Bonnie. As you said, I'm not actually confident in myself at all, but it's mostly bc of you, who came and filled in the parts of me that I couldn't do at all, who came and filled in the parts that I didn't want to do at all in the beginning bc I was too afraid. It's bc of you that I am who I am today. If it weren't you, then it couldn't be anyone else. Thank you so much for always helping me in everything. For our show today, even though it's our second time, at first I was like, "It's going to be chill. Of course we can do it. We could do a full hour concert with just the two of us, this will pass in no time, we can do it!" But it turned out harder than every other time for many reasons. We had many things to do too. But I was willing to do everything and I was honestly very happy during the show. I feel like I don't usually get to see myself let myself go, let loose, and just chill to this extent. I really enjoyed this time and that I was brave enough to try many new things. 🦊: Thank you, Bonnie, and thank you to all my friends here for having fun together. This time, before going on stage, the feeling was like, "Alright! Let's do it, let's get on stage!" It felt really chill. Usually, I put a lot of pressure on myself. I'm glad that this time, I have a lot of friends here to help fill in the things that in the past I never thought I would've been able to do. Thank you so much, everyone. Thank you for having fun and letting loose together. 🦊: I'd also like to thank the executives for always giving me opportunities. Like what others have said, I don't feel that confident at all, but when I get opportunities and the executives believe that I can do it, it really makes me feel more confident. Thank you so much for giving me many opportunities in many different things and for giving them to me since the start. Truly, thank you so much. 🦊: Thank you as well to the crew. Actually, things behind the scenes are very hectic. There are a lot of things that have to be done this time, and they've all been very occupied taking care of us. Thank you for taking care of us and supporting us, both the staff in front of and behind the scenes, including the control technicians. 🦊: Lastly, thank you to all the fans. Without you all since day one, we wouldn't be here today as the Blush Blossom girls. Truly, thank you so much for supporting us even if you didn't know what this was going to be. Ever since the first time you've supported us, we've grown and expanded and gotten this big now. Thank you so much. 🦊: And thank you to myself as well. I'm so glad that I was brave enough to do this. Thank you.

K-bab

16,503 просмотров • 2 месяцев назад

UPDATE: Pete Z and I are half way finished filming the second NDIS video and we will hopefully have it out to you all by the end of April. I just want to thank everybody again so much for their support. This has been by far the most amount of support I have ever received for anything. I was plugging away at my political journalism and activism work for about 6 or 7 hours a day every day for 6 years with not much to show for it. There were dark times back in 2023 when I had literally $50 in my account and I had to ask my parents for loans to buy fuel for my car. Now for the first time in years I can actually live without financial stress. I’m able to finally make this my full time professional job. And I’m only able to do that because of the massive amount of support I’ve had from the public. Like 300,000 extra followers on all my accounts in just one month. Thank you so much everybody. I want to spend my life fighting for a better Australia and fighting for the future of the West. I feel so blessed that I finally have support, thank you so much.
2:08

UPDATE: Pete Z and I are half way finished filming the second NDIS video and we will hopefully have it out to you all by the end of April. I just want to thank everybody again so much for their support. This has been by far the most amount of support I have ever received for anything. I was plugging away at my political journalism and activism work for about 6 or 7 hours a day every day for 6 years with not much to show for it. There were dark times back in 2023 when I had literally $50 in my account and I had to ask my parents for loans to buy fuel for my car. Now for the first time in years I can actually live without financial stress. I’m able to finally make this my full time professional job. And I’m only able to do that because of the massive amount of support I’ve had from the public. Like 300,000 extra followers on all my accounts in just one month. Thank you so much everybody. I want to spend my life fighting for a better Australia and fighting for the future of the West. I feel so blessed that I finally have support, thank you so much.

Drew Pavlou 🇦🇺🇺🇸🇺🇦🇹🇼

30,953 просмотров • 4 месяцев назад

There's so much focus on "how can AI do my work for me?" I think the more important question is "what work can I now do with AI that I would have never attempted before?" Earlier this year I wrote freestiler, a vector tiling engine for R and Python, with the help of Claude and Codex. I knew what the ideal engine looked like and how it would work at a high level. I didn't know how to put it together, and I don't know Rust, the language I wanted under the hood. Previously I would never have attempted this project as the ROI wasn't there. It would have taken me a year or more to learn the internals of a vector tiling engine and enough Rust to implement one. With Opus-level models, I could take it on. freestiler now powers all my vector tiling pipelines, including the map below rendering 143 million jobs from LODES, and it has 114 GitHub stars. Building this way has required a different set of skills. I don't review the code line by line. I set up adversarial agents to do that and write the test suites. What I review is the architecture, the behavior, and the results. Agent teams surface findings and explain their reasoning; I evaluate and critique. My job isn't to stress over code formatting, but instead to focus on questions like whether the engine is designed right, whether the output is correct, and if the UX makes sense. This means that I haven't "replaced my work." I've taken on entirely new work, with the help of agents, that I would have never done otherwise. It has taken some getting used to shipping code I haven't personally typed. In the old way of working, I built understanding through writing that code. Now I build understanding through managing the project - writing a spec, reviewing structure, evaluating UX. And that's helped me think a whole lot bigger in terms of what I can now do.
1:00

There's so much focus on "how can AI do my work for me?" I think the more important question is "what work can I now do with AI that I would have never attempted before?" Earlier this year I wrote freestiler, a vector tiling engine for R and Python, with the help of Claude and Codex. I knew what the ideal engine looked like and how it would work at a high level. I didn't know how to put it together, and I don't know Rust, the language I wanted under the hood. Previously I would never have attempted this project as the ROI wasn't there. It would have taken me a year or more to learn the internals of a vector tiling engine and enough Rust to implement one. With Opus-level models, I could take it on. freestiler now powers all my vector tiling pipelines, including the map below rendering 143 million jobs from LODES, and it has 114 GitHub stars. Building this way has required a different set of skills. I don't review the code line by line. I set up adversarial agents to do that and write the test suites. What I review is the architecture, the behavior, and the results. Agent teams surface findings and explain their reasoning; I evaluate and critique. My job isn't to stress over code formatting, but instead to focus on questions like whether the engine is designed right, whether the output is correct, and if the UX makes sense. This means that I haven't "replaced my work." I've taken on entirely new work, with the help of agents, that I would have never done otherwise. It has taken some getting used to shipping code I haven't personally typed. In the old way of working, I built understanding through writing that code. Now I build understanding through managing the project - writing a spec, reviewing structure, evaluating UX. And that's helped me think a whole lot bigger in terms of what I can now do.

Kyle Walker

13,740 просмотров • 1 месяц назад

Adobe.. has released a tool that combines Generative Gaussian Splats with a Diffusion layer and it's not all over the internet? WHAT IS GOING ON :D I had to test this out ofc! Substance 3D viewer, the new 3d viewer just released by Adobe, not only supports viewing of a large, large list of 3d formats, but it also supports generating of 3d models as Gaussian Splats. Combine that with the built-in 3d to image functionality and you basically have at least a few of the steps I've been doing for the last year or so using multiple tools. Exciting stuff! The new Photoshop Beta also supports the 3D viewer and you can import 3d files directly in photoshop as smart objects linked to the viewer. The elements you prompt in based on your models, can be easily exported to your clipboard without backgrounds making it very easy to use this in your normal image editing workflows. I have a suspicion that we'll see much more of this in other Adobe tools and I'm very curious to see this used for things like this in more 3d tools than "just" Adobe Neo (which is fantastic btw). #adobe #art #gaussiansplatting
0:38

Adobe.. has released a tool that combines Generative Gaussian Splats with a Diffusion layer and it's not all over the internet? WHAT IS GOING ON :D I had to test this out ofc! Substance 3D viewer, the new 3d viewer just released by Adobe, not only supports viewing of a large, large list of 3d formats, but it also supports generating of 3d models as Gaussian Splats. Combine that with the built-in 3d to image functionality and you basically have at least a few of the steps I've been doing for the last year or so using multiple tools. Exciting stuff! The new Photoshop Beta also supports the 3D viewer and you can import 3d files directly in photoshop as smart objects linked to the viewer. The elements you prompt in based on your models, can be easily exported to your clipboard without backgrounds making it very easy to use this in your normal image editing workflows. I have a suspicion that we'll see much more of this in other Adobe tools and I'm very curious to see this used for things like this in more 3d tools than "just" Adobe Neo (which is fantastic btw). #adobe #art #gaussiansplatting

Martin Nebelong

280,050 просмотров • 1 год назад

🐰 i worked a lot with El Guincho. I got to listen to a lot of tracks and i was really wowed by just how fresh and new this was. One of the first thoughts i had was ‘wow, you can approach a song like this?’ and i felt a lot of pressure about how to make the best use of it. so that must be why i wanted to stay close, it was a totally a different vibe, a whole new vibe. i hope to work again and have another session with the producers and songwriters that we worked with
0:52

🐰 i worked a lot with El Guincho. I got to listen to a lot of tracks and i was really wowed by just how fresh and new this was. One of the first thoughts i had was ‘wow, you can approach a song like this?’ and i felt a lot of pressure about how to make the best use of it. so that must be why i wanted to stay close, it was a totally a different vibe, a whole new vibe. i hope to work again and have another session with the producers and songwriters that we worked with

jk.97

26,798 просмотров • 5 месяцев назад

After starting the year with 2 top 10s to qualify for the Arnold Palmer and The Players, Joel Dahmen missed both those cuts and had to get an Uber to Monday Qualifying for the Valspar. He missed at qualifying, but got in the field anyway as an alternate and now has shot rounds of 72 and 70 to make the cut. After the round, he spoke about how grounding life on the PGA Tour can sometimes be: “Look, life on the PGA Tour is amazing. Wouldn't trade it for anything but also people probably don't see behind the scenes as much and so I got the family on the road, which is great, but it also can be difficult when you drive across the state on Sunday, get into a new place, you have to unpack, get them all ready, got to get groceries in the house, all those little things, thankfully my wife does most of that. “But then I had a 9:30 tee time and it was about an hour away and we didn't have another car. So I didn't have time to grab another one. So it was just Uber was the best option and yeah, so it's a lot different from, you know, the steak and wine and The Players clubhouse and an amazing week to driving to Brooksville to Monday. “And I hadn't had to do that in a while, thankfully, but the game humbles you quickly but also can recenter you and ground you and, yeah, I mean I didn't play well there on Monday either, but you knew I had a good chance of getting in the field and so it was nice little reset to hit some striped range balls on not a perfect golf course and realize how lucky we are to be out here on tour.” He’s currently 65th on the FedEx Cup Standings and will be hoping to go low this weekend and improve on that.
1:41

After starting the year with 2 top 10s to qualify for the Arnold Palmer and The Players, Joel Dahmen missed both those cuts and had to get an Uber to Monday Qualifying for the Valspar. He missed at qualifying, but got in the field anyway as an alternate and now has shot rounds of 72 and 70 to make the cut. After the round, he spoke about how grounding life on the PGA Tour can sometimes be: “Look, life on the PGA Tour is amazing. Wouldn't trade it for anything but also people probably don't see behind the scenes as much and so I got the family on the road, which is great, but it also can be difficult when you drive across the state on Sunday, get into a new place, you have to unpack, get them all ready, got to get groceries in the house, all those little things, thankfully my wife does most of that. “But then I had a 9:30 tee time and it was about an hour away and we didn't have another car. So I didn't have time to grab another one. So it was just Uber was the best option and yeah, so it's a lot different from, you know, the steak and wine and The Players clubhouse and an amazing week to driving to Brooksville to Monday. “And I hadn't had to do that in a while, thankfully, but the game humbles you quickly but also can recenter you and ground you and, yeah, I mean I didn't play well there on Monday either, but you knew I had a good chance of getting in the field and so it was nice little reset to hit some striped range balls on not a perfect golf course and realize how lucky we are to be out here on tour.” He’s currently 65th on the FedEx Cup Standings and will be hoping to go low this weekend and improve on that.

Flushing It

480,728 просмотров • 5 месяцев назад

Karpathy said something you'll regret ignoring: "We have to keep the AI on the leash. I'm still the bottleneck. I have to make sure this thing isn't introducing bugs and that there's no security issues." He said it at YC talk last year, when the worry was reliability. The models hallucinated and made mistakes no human would, so the leash implied keeping yourself in the loop and checking the output before trusting it. The models are far better now, and the line still holds, for a reason he was not focused on back then. Even a model that writes flawless code today still has no idea who is allowed to run it. Correctness and authorization are different problems, and only correctness improves as the model improves. A perfect agent still hands a tool where anyone can do anything, because permission was never part of the task. I actually tested this in practice with Claude Code. I asked it to build a small internal tool with a button that issues account credits. It worked first try, and running it locally, the credit applied the instant I clicked. Nothing decided who was allowed to click it. The agent wrote the right logic and displayed a success notification. It never checked whether the caller had the right, whether it should pause for a human, or whether anything was logged. And this is not a bug a smarter model can outgrow because the leash was never in the code. Identity, permissions, and audit live in the system that runs the app, not in what the agent generates. To solve this, I took the exact same bundle and hosted it on Retool. The credit write that fired silently on my laptop now stopped at an approval gate, resolved to a real identity through SSO, and landed in an audit log. I wrote none of it. The app inherited the entire boundary the moment it was deployed, and the video shows the before and after. You can try it yourself here: I also wrote a detailed breakdown of the whole thing in my recent article, and I worked with the team to put this together. It walks through the build, the exact moment the credit write went through on my laptop with nobody checking, and then what changed when the same app ran on Retool. It also covers why this is a property of the runtime and not something a better model fixes, which is why devs typically miss this. The article is quoted below.
0:38

Karpathy said something you'll regret ignoring: "We have to keep the AI on the leash. I'm still the bottleneck. I have to make sure this thing isn't introducing bugs and that there's no security issues." He said it at YC talk last year, when the worry was reliability. The models hallucinated and made mistakes no human would, so the leash implied keeping yourself in the loop and checking the output before trusting it. The models are far better now, and the line still holds, for a reason he was not focused on back then. Even a model that writes flawless code today still has no idea who is allowed to run it. Correctness and authorization are different problems, and only correctness improves as the model improves. A perfect agent still hands a tool where anyone can do anything, because permission was never part of the task. I actually tested this in practice with Claude Code. I asked it to build a small internal tool with a button that issues account credits. It worked first try, and running it locally, the credit applied the instant I clicked. Nothing decided who was allowed to click it. The agent wrote the right logic and displayed a success notification. It never checked whether the caller had the right, whether it should pause for a human, or whether anything was logged. And this is not a bug a smarter model can outgrow because the leash was never in the code. Identity, permissions, and audit live in the system that runs the app, not in what the agent generates. To solve this, I took the exact same bundle and hosted it on Retool. The credit write that fired silently on my laptop now stopped at an approval gate, resolved to a real identity through SSO, and landed in an audit log. I wrote none of it. The app inherited the entire boundary the moment it was deployed, and the video shows the before and after. You can try it yourself here: I also wrote a detailed breakdown of the whole thing in my recent article, and I worked with the team to put this together. It walks through the build, the exact moment the credit write went through on my laptop with nobody checking, and then what changed when the same app ran on Retool. It also covers why this is a property of the runtime and not something a better model fixes, which is why devs typically miss this. The article is quoted below.

Akshay 🚀

42,911 просмотров • 2 месяцев назад

A message from Catherine, The Princess of Wales As the summer comes to an end, I cannot tell you what a relief it is to have finally completed my chemotherapy treatment. The last nine months have been incredibly tough for us as a family. Life as you know it can change in an instant and we have had to find a way to navigate the stormy waters and road unknown. The cancer journey is complex, scary and unpredictable for everyone, especially those closest to you. With humility, it also brings you face to face with your own vulnerabilities in a way you have never considered before, and with that, a new perspective on everything. This time has above all reminded William and me to reflect and be grateful for the simple yet important things in life, which so many of us often take for granted. Of simply loving and being loved. Doing what I can to stay cancer free is now my focus. Although I have finished chemotherapy, my path to healing and full recovery is long and I must continue to take each day as it comes. I am however looking forward to being back at work and undertaking a few more public engagements in the coming months when I can. Despite all that has gone before I enter this new phase of recovery with a renewed sense of hope and appreciation of life. William and I are so grateful for the support we have received and have drawn great strength from all those who are helping us at this time. Everyone’s kindness, empathy and compassion has been truly humbling. To all those who are continuing their own cancer journey – I remain with you, side by side, hand in hand. Out of darkness, can come light, so let that light shine bright. 🎬 by Will Warr
3:06

A message from Catherine, The Princess of Wales As the summer comes to an end, I cannot tell you what a relief it is to have finally completed my chemotherapy treatment. The last nine months have been incredibly tough for us as a family. Life as you know it can change in an instant and we have had to find a way to navigate the stormy waters and road unknown. The cancer journey is complex, scary and unpredictable for everyone, especially those closest to you. With humility, it also brings you face to face with your own vulnerabilities in a way you have never considered before, and with that, a new perspective on everything. This time has above all reminded William and me to reflect and be grateful for the simple yet important things in life, which so many of us often take for granted. Of simply loving and being loved. Doing what I can to stay cancer free is now my focus. Although I have finished chemotherapy, my path to healing and full recovery is long and I must continue to take each day as it comes. I am however looking forward to being back at work and undertaking a few more public engagements in the coming months when I can. Despite all that has gone before I enter this new phase of recovery with a renewed sense of hope and appreciation of life. William and I are so grateful for the support we have received and have drawn great strength from all those who are helping us at this time. Everyone’s kindness, empathy and compassion has been truly humbling. To all those who are continuing their own cancer journey – I remain with you, side by side, hand in hand. Out of darkness, can come light, so let that light shine bright. 🎬 by Will Warr

The Prince and Princess of Wales

9,053,746 просмотров • 1 год назад

jungkook on “hooligan” and working with “el guincho”! 🐰: i worked a lot with “el guincho” 🐿️: jungkook worked a lot on the song 'Hooligan' 🐰: right! i got to listen to a lot of tracks and I was really wowed by just how fresh and new it was. one of the first thoughts i had was 'wow, you can approach a song like this?' and i felt a lot of pressure about how to make the best use of it. so that must be why i wanted to stay close. so, anyway... it was a totally different vibe. a whole new vibe. i hope to work again and have another session with the producers and the songwriters that we worked with.
0:58

jungkook on “hooligan” and working with “el guincho”! 🐰: i worked a lot with “el guincho” 🐿️: jungkook worked a lot on the song 'Hooligan' 🐰: right! i got to listen to a lot of tracks and I was really wowed by just how fresh and new it was. one of the first thoughts i had was 'wow, you can approach a song like this?' and i felt a lot of pressure about how to make the best use of it. so that must be why i wanted to stay close. so, anyway... it was a totally different vibe. a whole new vibe. i hope to work again and have another session with the producers and the songwriters that we worked with.

ajk ⋆

95,476 просмотров • 5 месяцев назад

I never understood why these aren’t more common in the states. In other countries I’ve been to they are more common or the standard. I feel the cut back on toilet paper and they really do leave you with a fresh out of the shower feeling, it’s a win win. While this woman’s reaction can be quite common, after the initial shock of how it works, most people actually enjoy it. It can be life changing for some and they never go back to toilet paper. I think anyone who has never tried one of these, should try it. I feel the benefits of them outweigh the cost. Most of the time retrofit kits are also available for more toilets as well. I remember the first time I ever used one as well, my first experience wasn’t as nice because the water was cold, but it has since been much better.
0:50

I never understood why these aren’t more common in the states. In other countries I’ve been to they are more common or the standard. I feel the cut back on toilet paper and they really do leave you with a fresh out of the shower feeling, it’s a win win. While this woman’s reaction can be quite common, after the initial shock of how it works, most people actually enjoy it. It can be life changing for some and they never go back to toilet paper. I think anyone who has never tried one of these, should try it. I feel the benefits of them outweigh the cost. Most of the time retrofit kits are also available for more toilets as well. I remember the first time I ever used one as well, my first experience wasn’t as nice because the water was cold, but it has since been much better.

SonnyBoy🇺🇸

29,965 просмотров • 3 месяцев назад

gunwook day 1 here&now encore ending ment 🥹 💌 the first day of the encore concert has ended. it’s coming to an end. since it’s been a while since we were together with zerose and meeting you again to perform like this, i was really nervous but also really happy. even while preparing, i felt very excited. there were things we were showing for the first time, and many stages we were performing for the first time with this set, so it was really fun. more than anything, the fact that the nine of us were building and carrying the stage together felt especially precious during this concert. since i’m one of the taller members in the team and i usually stand at the very end in the official formations, i get a lot of chances to see all the members at once. every time that happens, i feel so reassured, and i really love being able to see the backs of our amazing members seeing the backs of the other eight members and also seeing their side profiles. i hope that from now on, for a long, long time, i can keep capturing those moments and those in my eyes, and i also hope i can keep filling my eyes with zerose’s blue lights and your blue smiles. somehow, i remember writing this in a letter back during our first anniversary. i said that i felt sad and upset that there might be an end, because i was so curious about what our future selves would look like. and i think i still feel that way now. i don’t know what will happen in the future, but i’ve become so curious about the futures of the other members too, and i really want to support those futures. so because of that, i think the feeling of sadness becomes a little bigger as time passes. for some zerose, today might be the last time you see us like this, or maybe you’ll see us again tomorrow, or the day after tomorrow. but i hope that you can take the image of the nine of us as zerobaseone and keep it in your hearts forever, treasuring it as something precious. and i hope we can become a presence you can always look back on something that gives you strength whenever you need it. thank you. i love you.
2:39

gunwook day 1 here&now encore ending ment 🥹 💌 the first day of the encore concert has ended. it’s coming to an end. since it’s been a while since we were together with zerose and meeting you again to perform like this, i was really nervous but also really happy. even while preparing, i felt very excited. there were things we were showing for the first time, and many stages we were performing for the first time with this set, so it was really fun. more than anything, the fact that the nine of us were building and carrying the stage together felt especially precious during this concert. since i’m one of the taller members in the team and i usually stand at the very end in the official formations, i get a lot of chances to see all the members at once. every time that happens, i feel so reassured, and i really love being able to see the backs of our amazing members seeing the backs of the other eight members and also seeing their side profiles. i hope that from now on, for a long, long time, i can keep capturing those moments and those in my eyes, and i also hope i can keep filling my eyes with zerose’s blue lights and your blue smiles. somehow, i remember writing this in a letter back during our first anniversary. i said that i felt sad and upset that there might be an end, because i was so curious about what our future selves would look like. and i think i still feel that way now. i don’t know what will happen in the future, but i’ve become so curious about the futures of the other members too, and i really want to support those futures. so because of that, i think the feeling of sadness becomes a little bigger as time passes. for some zerose, today might be the last time you see us like this, or maybe you’ll see us again tomorrow, or the day after tomorrow. but i hope that you can take the image of the nine of us as zerobaseone and keep it in your hearts forever, treasuring it as something precious. and i hope we can become a presence you can always look back on something that gives you strength whenever you need it. thank you. i love you.

𐙚

20,944 просмотров • 5 месяцев назад

I've migrated the old Mast3r-SLAM example I had made last year to the latest version of Rerun and made a bunch of improvements! I wanted to spend some time with agents to modernize it. Here's an example of me walking around with my iPhone and getting a dense reconstruction at about 10FPS on a 5090. Heres the following improvements I made. Brought it into the monorepo with proper packaging: • Using prefix.dev pixi-build to get rid of all the mast3r/asmk/lietorch vendored code with just a few small patches. This let me remove so 60k lines of code from the repo! • Don't have to build the lietorch code on my machine anymore, which was taking ~10 minutes to compile (and also made it work on blackwell when it previously did not) Rebuilt the Gradio interface: • Fixed incremental updates, .MOV uploads, and stop behavior • Made the CLI + Gradio interface share the same entry point so updates automatically propagate Upgraded the Rerun integration: • Switched to a multiprocessing async logging strategy • Added video/pointmap/confidence logging • Improved blueprint layout and hid noisy entities from 3D view • Biggest perf win was the async background logger - documented about a ~2.5x speedup from decoupling logging from tracking The newest and most interesting part was my attempt to replace the CUDA kernels for Gauss-Newton ray matching with a Modular Mojo backend. As a Python dev, every time I look at CUDA code I basically shy away as it's pretty difficult for me to understand. Mojo let me rewrite the matching logic in a syntax I'm more comfortable with while still getting near-CUDA performance. Mojo is now the default matching backend with CUDA fallback. One major piece that's missing is the custom PyTorch op path, but I'll eventually do that as well. I heavily leaned on Claude Code to do the CUDA → Mojo migration, and I have no doubt it's not the cleanest or most idiomatic, BUT it's way more readable for me and helps me better understand the underlying algorithm. This was a ton of work, and a large part of why I'm doing it is how the monorepo compounds. This becomes an artifact for the next example I want to build with Claude that I can point to, which will make it even faster to implement. The compounding nature of this is really interesting and part of why I'm spending so much time trying to make things nice and readable.
0:21

I've migrated the old Mast3r-SLAM example I had made last year to the latest version of Rerun and made a bunch of improvements! I wanted to spend some time with agents to modernize it. Here's an example of me walking around with my iPhone and getting a dense reconstruction at about 10FPS on a 5090. Heres the following improvements I made. Brought it into the monorepo with proper packaging: • Using prefix.dev pixi-build to get rid of all the mast3r/asmk/lietorch vendored code with just a few small patches. This let me remove so 60k lines of code from the repo! • Don't have to build the lietorch code on my machine anymore, which was taking ~10 minutes to compile (and also made it work on blackwell when it previously did not) Rebuilt the Gradio interface: • Fixed incremental updates, .MOV uploads, and stop behavior • Made the CLI + Gradio interface share the same entry point so updates automatically propagate Upgraded the Rerun integration: • Switched to a multiprocessing async logging strategy • Added video/pointmap/confidence logging • Improved blueprint layout and hid noisy entities from 3D view • Biggest perf win was the async background logger - documented about a ~2.5x speedup from decoupling logging from tracking The newest and most interesting part was my attempt to replace the CUDA kernels for Gauss-Newton ray matching with a Modular Mojo backend. As a Python dev, every time I look at CUDA code I basically shy away as it's pretty difficult for me to understand. Mojo let me rewrite the matching logic in a syntax I'm more comfortable with while still getting near-CUDA performance. Mojo is now the default matching backend with CUDA fallback. One major piece that's missing is the custom PyTorch op path, but I'll eventually do that as well. I heavily leaned on Claude Code to do the CUDA → Mojo migration, and I have no doubt it's not the cleanest or most idiomatic, BUT it's way more readable for me and helps me better understand the underlying algorithm. This was a ton of work, and a large part of why I'm doing it is how the monorepo compounds. This becomes an artifact for the next example I want to build with Claude that I can point to, which will make it even faster to implement. The compounding nature of this is really interesting and part of why I'm spending so much time trying to make things nice and readable.

Pablo Vela

42,143 просмотров • 4 месяцев назад