Apoush

Most crypto products still stop at holding value You keep stablecoins in a wallet maybe move them between apps maybe send them to an exchange then wait for the real world part to begin That gap is the problem The interesting thing about KAST is that it does not treat stablecoins as something that should stay trapped inside crypto rails It turns them into something closer to everyday money You can store digital dollars receive global payments earn on idle balances move money across borders and spend through a card when life actually asks for payment That full flow matters more than one single feature Because most people do not wake up thinking about financial infrastructure They just want money that works when they need it @KASTxyz feels important because it is trying to make stablecoins useful beyond trading and holding not louder just more usable

@KASTxyz 👀

Matchday is coming ⚽ Are you ready?

@AIExplainedHQ @0xsikdar @MohdSarim0 Yea Bro

@apoushDiv @0xsikdar @MohdSarim0 Exactly, robots trained only on perfect conditions usually fall apart when reality gets messy

Lab data is useful, but it usually shows the clean version of the world. The real world is different. Objects are not always in the right place, lighting changes, hands move imperfectly, rooms are cluttered, and people do things in ways that are hard to script. That mess is not a problem for robotics data. It is the point. @RoboX_to matters because it helps capture the kind of first person data robots actually need if they are going to work outside controlled environments. @0xsikdar , @0xnag1 , @MohdSarim0

@RoboX_to Now even drinking water can be valuable data who would’ve thought!?

EgoTask is a data collection campaign for robotics training. First-person video of everyday manipulation tasks, recorded on an iPhone or GoPro, one to five minutes per session. During recording, RoboX tracks both hands and labels every grasp, then turns the footage into training data for imitation learning: hand poses, grasp types, camera trajectory, and object contacts.

@0xsikdar @MohdSarim0 Thanks Ser 🩷

@apoushDiv @MohdSarim0 good one i like it

@0x_obscuro @0xsikdar @MohdSarim0 gRoboX :))

@apoushDiv @0xsikdar @MohdSarim0 Broooooooo just ROBOX

@MohdSarim0 Tnx Broski 🩷🫂

@apoushDiv just amazing

One of the biggest gaps in robotics is not that teams do not know what they want to build, it is that they do not have enough real world data to train it properly. A robot that only learns from clean lab setups will struggle when the room is messy, the object is slightly different, or the movement does not go as planned. @RoboX_to is interesting because it connects normal people with robotics teams through useful first person recordings. People capture real tasks from real environments, and that data can become training material for robots that need to understand the world outside the lab. @0xsikdar , @0xnag1 ,@MohdSarim0

@rehmanorg @Camilee_KAST gKast Bro

KAST has introduced a limited-time collaboration with Pudgy Penguins featuring a giveaway for the Pengu Black Card (Metal Edition), valued at $1000. Users who want to participate must complete identity verification, join the Pengu Card waitlist, and fulfill a minimum $100 requirement through deposit or spending. The process is open to both new and existing users, with different conditions depending on account type. After completing all steps, users are included in the selection pool where winners are picked periodically. Notifications will be shared through the @KASTxyz app and email. This campaign offers access to a premium card tier with limited availability and scheduled winner announcements.

@Mlvmahdii @axisrobotics @BitRobotNetwork gBot!

Huge congrats to @axisrobotics and @BitRobotNetwork on joining forces! The synergy here is unmatched, and I'm incredibly hyped to see what you guys cook up next. Let's build the future together!

@Mlvmahdii @PrismaXai @shayebackus @vivianrobotics @MaxC16134 gMic Broski

Gprisma my friends No spotlight. No applause. Just thousands of teleoperated actions turning human skill into robotic intelligence. One day it's training. The next day it's autonomy. @PrismaXai @shayebackus @vivianrobotics @MaxC16134

@0xsikdar @RoboX_to gRoboX

Do you know what the egocentric view means, and why is it more reliable data for training robots? Well, in @RoboX_to, we collect only egocentric data, and that is for a reason. Coming soon...

@goo_vision Aww! :((

Prickly 🌵

@ruthwikdasyam @dimensionalos Nice Pet!

Holy shit.. SF’s full of robotic dogs.. People at @dimensionalos are mapping all over San Francisco

@abhihereandnow again Try!

Went to a robot fight last night, not a single punch in sight.

@lukas_m_ziegler Rubik's game?

Robot hands are getting out of hand! ✊🏼 RGB-only dexterous hand control via 3D Gaussian splatting. ETH Zürich researchers developed ViserDex that is a sim-to-real framework for in-hand object reorientation using only a monocular RGB camera and 3D Gaussian Splatting. So in this case, the domain randomization happens in Gaussian representation space, not pixel space. They apply physically consistent augmentations directly to 3D Gaussians like shadows, marks, spectral effects, lighting shifts, generating photorealistic visual data from a single static 3D model. A monocular pose estimator trained on this Gaussian-generated data predicts object pose. A manipulation policy trained with curriculum-based RL and teacher-student distillation converts those poses into dexterous hand control. What do we get from that? Zero-shot transfer to an Allegro multi-fingered hand. Successfully reoriented five diverse objects: cubes, bottles, ducks, globes, 3D-printed toys. Over 25 consecutive successful reorientations. Read the paper here: rffr.leggedrobotics.com/works/viserdex/ ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com

@_jhunsaker Ooomm 😋

won't be able to move shortly

@ronitcryptoOG @0xsikdar @MohdSarim0 @RoboX_to 😍

@apoushDiv @0xsikdar @MohdSarim0 @RoboX_to wow

Robotics does not only have a model problem, it has a data problem. Language models had the internet to learn from, image models had billions of pictures, but robots need something much harder: real movement, real rooms, real hands, real objects, and real mistakes. That is the part @RoboX_to is trying to solve by turning normal smartphones into first person data sensors for robotics training. Not inside a perfect lab, but inside the messy world robots actually need to understand. @0xsikdar , @0xnag1 , @MohdSarim0