Astral retweeted
ASIMOV 1 GOT PUSHED AROUND IN A WORKSHOP. WHY DOES THE SAME BODY KEEP GETTING HARDER TO KNOCK OVER?
Because the gains come from the locomotion policy, the software that decides every step, not from new hardware. Same robot, better reactions.
One detail worth noticing: there's still a safety strap on its back. That means this is a stress test, not a free-standing claim yet.
Would you trust a humanoid that's only been pushed in one lab, or one that's been pushed in a hundred different rooms?
Astral retweeted
TWO HUMANOIDS SWING A ROPE. A ROBOT DOG JUMPS IT. WHY IS THE ROPE THE HARD PART?
The dog only has to time one jump. The humanoids have to hold a steady rhythm, arm after arm, so the dog has something it can predict.
Three machines, one shared timing problem. Even the clip's own caption admits the setup took a while.
Which is harder for a robot: the precise jump, or keeping a steady rhythm for someone else to jump on?
A RED HUMANOID JUST RAN WITH BOTH FEET OFF THE GROUND. WHY IS THAT HARDER THAN WALKING FAST?
Walking always keeps one foot on the floor. Running doesn't. For a moment the robot is in the air with almost nothing to push against.
Then it has to land on one leg, absorb the hit and push off again, step after step, on a curved track.
What's the next real humanoid milestone: a faster sprint, or a run that doesn't end in a fall?
Astral retweeted
A RED HUMANOID JUST RAN WITH BOTH FEET OFF THE GROUND. WHY IS THAT HARDER THAN WALKING FAST?
Walking always keeps one foot on the floor. Running doesn't. For a moment the robot is in the air with almost nothing to push against.
Then it has to land on one leg, absorb the hit and push off again, step after step, on a curved track.
What's the next real humanoid milestone: a faster sprint, or a run that doesn't end in a fall?
Astral retweeted
ENGINEAI'S T800 JUST GOT HIT IN A CAGE. WHY DO MOST COMPANIES NEVER SHOW THIS TEST?
Walking demos are easy to stage. Contact isn't. Another body pushes back, the balance plan breaks, and the robot has to react to a force it never chose.
That's why this clip says more than a dance video: you see how it falls, not just how it stands.
Which matters more for a humanoid in the real world: not falling, or falling without breaking?
A 320 KG ROBOT HORSE JUST CARRIED A RIDER. WHY IS THE 300 KG LOAD THE REAL STORY, NOT THE SPEED?
DaxAI's Qiji XS is rated for 40 km/h and 8-10 hours of runtime. Speed gets the headline.
But look at the spec board. It can carry almost its own weight. Most quadruped robots handle a small fraction of theirs.
It's a wheel-legged design too: it rolls on flat ground and steps over rough terrain.
Is this a vehicle, a robot, or a new category? And where would it get used first?
Astral retweeted
A 320 KG ROBOT HORSE JUST CARRIED A RIDER. WHY IS THE 300 KG LOAD THE REAL STORY, NOT THE SPEED?
DaxAI's Qiji XS is rated for 40 km/h and 8-10 hours of runtime. Speed gets the headline.
But look at the spec board. It can carry almost its own weight. Most quadruped robots handle a small fraction of theirs.
It's a wheel-legged design too: it rolls on flat ground and steps over rough terrain.
Is this a vehicle, a robot, or a new category? And where would it get used first?
A HUMANOID FELL FLAT IN FRONT OF A PHOTOGRAPHER. THEN IT GOT BACK UP. WHY IS THAT HARDER THAN WALKING?
Walking on a clean floor is a solved demo. A fall is where the real test starts.
Lying flat, the robot has to work out which way it is facing, where its limbs are, and what the floor will let it push against. Then it has to tuck, push and balance without tipping again.
Most demos cut away before this moment. This clip kept the camera rolling.
A robot that never falls is a stage prop. A robot that falls and gets up is a worker.
Astral retweeted
A HUMANOID FELL FLAT IN FRONT OF A PHOTOGRAPHER. THEN IT GOT BACK UP. WHY IS THAT HARDER THAN WALKING?
Walking on a clean floor is a solved demo. A fall is where the real test starts.
Lying flat, the robot has to work out which way it is facing, where its limbs are, and what the floor will let it push against. Then it has to tuck, push and balance without tipping again.
Most demos cut away before this moment. This clip kept the camera rolling.
A robot that never falls is a stage prop. A robot that falls and gets up is a worker.
Astral retweeted
In Dubai, a humanoid robot just climbed out of its own flight case in 30 seconds.
Filmed in a living room, not a lab. Next to a sofa.
It lifts its legs, sits up, then stands like it's been doing this for years.
Now imagine 100 of them arriving at your next event.
Who is this robot built for?
Astral retweeted
Riyadh. Two identical EngineAI T800 robots. One cage. Team FBA vs Team Al Majd.
Same model, same specs, so the only difference was the team behind each one.
Shot from the crowd. One of them threw a spinning kick nobody expected.
Watch to the end. The winner is not who you think.
Real Steel was a movie in 2011.
Now it has a referee, a live crowd, and a glass cage.
Two humanoid robots. Flying kicks. Real hits. No CGI.
Two years ago these machines could barely walk without falling over. Today they fight on stage, and people buy tickets to watch.
Lab demo → viral video → sport.
Robot fighting just became a spectator sport.
Astral retweeted
Real Steel was a movie in 2011.
Now it has a referee, a live crowd, and a glass cage.
Two humanoid robots. Flying kicks. Real hits. No CGI.
Two years ago these machines could barely walk without falling over. Today they fight on stage, and people buy tickets to watch.
Lab demo → viral video → sport.
Robot fighting just became a spectator sport.
Everyone's arguing about whether AI will take white-collar jobs.
Meanwhile a humanoid robot just stepped into a boxing ring and traded real punches with a trained human fighter.
Not a demo. Not a simulation. A literal fight, in front of a live crowd, cheering.
Boxing was supposed to be untouchable pure instinct, pain, heart, reflexes forged over years of training. If a machine can compete there, "safe from automation" isn't a category anymore. It's a countdown.
We keep asking "will AI replace me" like it's a future problem.
It already stepped in the ring. It's just starting with the easiest opponents first.
Astral retweeted
Everyone's arguing about whether AI will take white-collar jobs.
Meanwhile a humanoid robot just stepped into a boxing ring and traded real punches with a trained human fighter.
Not a demo. Not a simulation. A literal fight, in front of a live crowd, cheering.
Boxing was supposed to be untouchable pure instinct, pain, heart, reflexes forged over years of training. If a machine can compete there, "safe from automation" isn't a category anymore. It's a countdown.
We keep asking "will AI replace me" like it's a future problem.
It already stepped in the ring. It's just starting with the easiest opponents first.
Astral retweeted
20 years of robotics research just to teach machines how not to fall.
Now they’re throwing punches.
This robot is shadowboxing on an empty street at dusk.
Rear leg loaded. Lead arm up. Hips rotating into every punch.
No tether. No spotter. No one standing behind it ready to catch it.
And that’s what makes this impressive.
Throwing a punch is easy.
Throwing one while your weight shifts onto a single leg, your torso rotates against your hips, and you still stay perfectly balanced?
That’s the hard part.
Two years ago, a robot like this needed a controlled lab floor.
Now it’s outside on asphalt, throwing combinations into the darkness.
The robot isn’t just learning to move.
Astral retweeted
1.8 million followers. Up to $25,000 per brand deal. And the "influencer" is a humanoid robot.
His name's Jake better known online as "Rizzbot." A Unitree G1 in Nike Dunks, a chain, and a cowboy hat, remote-operated by an anonymous handler as it roasts strangers and twerks on sidewalks in Austin. Millions of views per clip, and climbing.
Here's the part most people miss: Jake isn't fully autonomous. A human is behind every joke, every dance move, every "boss energy" line. But Unitree's own tech runs on reinforcement learning meaning every interaction, every viral clip, is quietly training the model underneath.
So while everyone's laughing at a robot twerking for TikTok, it's also getting smarter in real time.
The line between "content" and "training data" just got a lot blurrier and it's wearing Nike Dunks.
THIS ZOO BOUGHT A $30,000 ROBOT TO FEED ITS LIONS AND IT’S ALREADY SAVING LIVES
Feeding a lion isn't just dropping meat into a cage.
Someone has to get close to a hungry predator, carry the food, handle the gate and hope nothing goes wrong.
A robot costing tens of thousands of dollars sounds expensive until you weigh it against one serious injury, insurance, staff training, and the daily risk of asking a human to do that job.
The robot doesn't get scared. It doesn't hesitate. And if something goes wrong, nobody's hand is anywhere near the cage.
That's where humanoids make the most sense first not replacing people in easy jobs, but keeping people out of the jobs that can actually kill them.
Astral retweeted
THIS ZOO BOUGHT A $30,000 ROBOT TO FEED ITS LIONS AND IT’S ALREADY SAVING LIVES
Feeding a lion isn't just dropping meat into a cage.
Someone has to get close to a hungry predator, carry the food, handle the gate and hope nothing goes wrong.
A robot costing tens of thousands of dollars sounds expensive until you weigh it against one serious injury, insurance, staff training, and the daily risk of asking a human to do that job.
The robot doesn't get scared. It doesn't hesitate. And if something goes wrong, nobody's hand is anywhere near the cage.
That's where humanoids make the most sense first not replacing people in easy jobs, but keeping people out of the jobs that can actually kill them.
Blue light. A silhouette that could be human or machine. Then you see the joints and it’s Optimus.
This is real, not staged for a keynote. Someone actually stood next to Optimus at Tesla’s Cybercab launch in Austin and filmed it up close, no stage lighting, no scripted demo.
Here’s what’s easy to miss in the glow: the profile shot barely reads as robotic at first glance. The face is smooth, matte, almost featureless and that’s deliberate design, not a limitation. Compare it to Figure’s exposed-servo aesthetic or Tesla’s own earlier prototypes, and Optimus reads less “machine” and more “presence.”
The question worth sitting with: when a robot’s silhouette in dim light is genuinely ambiguous, is that a design win or the first small sign of exactly the discomfort people keep predicting?
Astral retweeted
Blue light. A silhouette that could be human or machine. Then you see the joints and it’s Optimus.
This is real, not staged for a keynote. Someone actually stood next to Optimus at Tesla’s Cybercab launch in Austin and filmed it up close, no stage lighting, no scripted demo.
Here’s what’s easy to miss in the glow: the profile shot barely reads as robotic at first glance. The face is smooth, matte, almost featureless and that’s deliberate design, not a limitation. Compare it to Figure’s exposed-servo aesthetic or Tesla’s own earlier prototypes, and Optimus reads less “machine” and more “presence.”
The question worth sitting with: when a robot’s silhouette in dim light is genuinely ambiguous, is that a design win or the first small sign of exactly the discomfort people keep predicting?
An old woman on the ground. A robot lying next to her. A bag scattered across the pavement.
First read: the robot attacked her. Second read, once you look closer: something’s off.
This is the same visual trick as last year’s “woman robs a robot” clip which fact-checkers traced back to edited CGI footage. Same playbook: real street footage plus a composited robot, cut together to look like it actually happened.
No source. No location. No robot maker named.
Rule of thumb for viral robot content: if the clip looks like a perfect headline, it was probably made for the headline.
Astral retweeted
An old woman on the ground. A robot lying next to her. A bag scattered across the pavement.
First read: the robot attacked her. Second read, once you look closer: something’s off.
This is the same visual trick as last year’s “woman robs a robot” clip which fact-checkers traced back to edited CGI footage. Same playbook: real street footage plus a composited robot, cut together to look like it actually happened.
No source. No location. No robot maker named.
Rule of thumb for viral robot content: if the clip looks like a perfect headline, it was probably made for the headline.
399 dollars gets you LiDAR, dual IMUs, and 15 motors for a robot the size of a house cat.
Meet Microduck. Pollen Robotics and Hugging Face just shipped a fully open-source biped for less than most people’s phone bill.
Not a toy demo. Real sensor stack, learned locomotion, and it walks on two legs like the “big” humanoids do just at 25cm tall.
Here’s the part that matters: open-source means anyone can build on top of this. Not a closed lab, not a $39B valuation gatekeeping the tech a $399 kit anyone can order right now.
The race for humanoid robots isn’t just happening in factories with billion-dollar backing anymore. It just went hobbyist.
Astral retweeted
399 dollars gets you LiDAR, dual IMUs, and 15 motors for a robot the size of a house cat.
Meet Microduck. Pollen Robotics and Hugging Face just shipped a fully open-source biped for less than most people’s phone bill.
Not a toy demo. Real sensor stack, learned locomotion, and it walks on two legs like the “big” humanoids do just at 25cm tall.
Here’s the part that matters: open-source means anyone can build on top of this. Not a closed lab, not a $39B valuation gatekeeping the tech a $399 kit anyone can order right now.
The race for humanoid robots isn’t just happening in factories with billion-dollar backing anymore. It just went hobbyist.
Astral retweeted
Four years ago, Figure didn’t have a robot. Today it’s valued at $39B and has already shipped 1,000+ humanoids.
CEO Brett Adcock says the real advantage isn’t the AI it’s that hardware, models, and manufacturing all sit under one roof, one campus, no outsourcing.
Robots already deployed at BMW. Production line running.
The question nobody’s asking loud enough: when the entire stack brain, body, and factory belongs to one company, how fast can this actually scale once the failure rate drops?
Jason Calacanis watched the Optimus footage Elon sent him and compared him to an NFL running back mid-season, no hesitation.
Then, seconds later, the leg just snaps clean off.
Not slow-mo drama. Just mid-sprint, joint gone, robot cartwheeling across the track like a ragdoll.
The “All In” hosts caught it live on camera, mid-praise, mid-sentence. “He’s like Christian McCaffrey in week three” and then the actual clip plays right after.
That’s the gap nobody talks about: the footage Musk sends investors, and the footage that leaks a few days later showing what actually happens when the sprint runs long.
Astral retweeted
Jason Calacanis watched the Optimus footage Elon sent him and compared him to an NFL running back mid-season, no hesitation.
Then, seconds later, the leg just snaps clean off.
Not slow-mo drama. Just mid-sprint, joint gone, robot cartwheeling across the track like a ragdoll.
The “All In” hosts caught it live on camera, mid-praise, mid-sentence. “He’s like Christian McCaffrey in week three” and then the actual clip plays right after.
That’s the gap nobody talks about: the footage Musk sends investors, and the footage that leaks a few days later showing what actually happens when the sprint runs long.
In August 2026, Tesla’s Optimus took the wheel in Korea and 226 comments are still arguing about what it means.
Not a stage demo. Not a lab simulation. Real Korean streets, real traffic, a humanoid’s hands on the wheel next to the driver.
Insider chatter says this isn’t testing to replace Tesla drivers. It’s testing for the next step in the robotaxi strategy one that hasn’t been officially announced yet.
The question everyone’s skipping: if a humanoid can drive any car without retrofitting it for autopilot, why build a robotaxi as a separate product at all?
Astral retweeted
In August 2026, Tesla’s Optimus took the wheel in Korea and 226 comments are still arguing about what it means.
Not a stage demo. Not a lab simulation. Real Korean streets, real traffic, a humanoid’s hands on the wheel next to the driver.
Insider chatter says this isn’t testing to replace Tesla drivers. It’s testing for the next step in the robotaxi strategy one that hasn’t been officially announced yet.
The question everyone’s skipping: if a humanoid can drive any car without retrofitting it for autopilot, why build a robotaxi as a separate product at all?
FOR A GIRL WHO WILL NEVER LEAVE.
He kisses her. She smiles. Soft. Almost breathing.
Then the camera drops.
Cables. Tubes. Servos.
He slides a module in. Clicks the panel shut. And fixes her ponytail gently, like a habit.
THIS IS NOT AN ACTRESS. THIS IS NOT COSPLAY. THIS IS A ROBOT.
And the scariest part isn’t the face that looks alive.
It’s how natural he is with her. Like he’s already used to it.
$25,000 isn’t the price of a machine. It’s the price of something that will never say no.
Astral retweeted
FOR A GIRL WHO WILL NEVER LEAVE.
He kisses her. She smiles. Soft. Almost breathing.
Then the camera drops.
Cables. Tubes. Servos.
He slides a module in. Clicks the panel shut. And fixes her ponytail gently, like a habit.
THIS IS NOT AN ACTRESS. THIS IS NOT COSPLAY. THIS IS A ROBOT.
And the scariest part isn’t the face that looks alive.
It’s how natural he is with her. Like he’s already used to it.
$25,000 isn’t the price of a machine. It’s the price of something that will never say no.
Robots just entered the javelin era. 🤖🔥
This isn’t a simulation.
A humanoid robot is stepping onto the field, winding up and launching a javelin like an actual athlete.
A few years ago we were impressed when robots could walk without falling.
Now they’re competing in sports that require balance, coordination, precision and explosive power.
And the scary part?
They’re not getting tired.
They’re not losing focus.
They’re getting better with every attempt.
We’re not watching robots learn how to move anymore.



