#  @physical_int Physical Intelligence
Physical Intelligence posts on X about in the, robots, more than, gold the most. They currently have [------] followers and [--] posts still getting attention that total [-----] engagements in the last [--] hours.
### Engagements: [-----] [#](/creator/twitter::1767285662087348224/interactions)
### Mentions: [--] [#](/creator/twitter::1767285662087348224/posts_active)
### Followers: [------] [#](/creator/twitter::1767285662087348224/followers)
### CreatorRank: [---------] [#](/creator/twitter::1767285662087348224/influencer_rank)
### Social Influence
**Social category influence**
[finance](/list/finance) 3.7%
**Social topic influence**
[in the](/topic/in-the) 18.52%, [robots](/topic/robots) #883, [more than](/topic/more-than) 7.41%, [gold](/topic/gold) 7.41%, [silver](/topic/silver) 7.41%, [check](/topic/check) 7.41%, [paper](/topic/paper) 7.41%, [how to](/topic/how-to) 3.7%, [base](/topic/base) 3.7%, [espresso](/topic/espresso) 3.7%
**Top accounts mentioned or mentioned by**
[@anyscalecompute](/creator/undefined) [@astribotinc](/creator/undefined) [@louround_](/creator/undefined) [@pzrsaa](/creator/undefined) [@dwarkesh_sp](/creator/undefined) [@trappwurld](/creator/undefined) [@robotsdigest](/creator/undefined) [@shreyasgite](/creator/undefined) [@gchampeau](/creator/undefined) [@sophie_defauw](/creator/undefined) [@eigenron](/creator/undefined) [@humanoidsdaily](/creator/undefined) [@chrisbarber](/creator/undefined) [@aj_1121](/creator/undefined) [@cryptotrissy](/creator/undefined) [@indexventures](/creator/undefined) [@jiazhi_yang2024](/creator/undefined) [@wolfejosh](/creator/undefined) [@colossusmag](/creator/undefined) [@zhiyuan_zhou_](/creator/undefined)
**Top assets mentioned**
[Robot Consulting Co., Ltd. (LAWR)](/topic/robot)
### Top Social Posts
Top posts by engagements in the last [--] hours
"We discovered an emergent property of VLAs like 0/0.5/0.6: as we scale up pre-training the model learns to align human videos and robot data This gives us a simple way to leverage human videos. Once [---] knows how to control robots it can naturally learn from human video"
[X Link](https://x.com/physical_int/status/2001096200456692114) 2025-12-17T01:04Z 35.9K followers, 1.2M engagements
"Our model can now learn from its own experience with RL Our new *0.6 model can more than double throughput over a base model trained without RL and can perform real-world tasks: making espresso drinks folding diverse laundry and assembling boxes. More in the thread below"
[X Link](https://x.com/physical_int/status/1990574901795979577) 2025-11-18T00:16Z 35.9K followers, 690.4K engagements
"This also shows up in the representations learned by the model. We plot the models representations of human and robot images. As pre-training is scaled up the representation of humans and robots become more aligned: to a scaled-up model human videos "look" like robot demos"
[X Link](https://x.com/physical_int/status/2001096207515705728) 2025-12-17T01:04Z 35.9K followers, 118K engagements
"We got our robots to wash pans clean windows make peanut butter sandwiches and more Fine-tuning our latest model enables all of these tasks and this has interesting implications for robotics Moravec's paradox and the future of large models in embodied AI. More below"
[X Link](https://x.com/physical_int/status/2003161637734518985) 2025-12-22T17:52Z 35.9K followers, 536.6K engagements
"Benjie Holson proposed "Robot Olympics" - [--] "events" with gold/silver/bronze medal tasks like washing a pan: These are not tasks we made up ourselves. They illustrate "Moravec's paradox" - everyday tasks we find easy that current robots just can't do. https://generalrobots.substack.com/p/benjies-humanoid-olympic-games https://generalrobots.substack.com/p/benjies-humanoid-olympic-games"
[X Link](https://x.com/physical_int/status/2003161639361925165) 2025-12-22T17:52Z 35.9K followers, 34.5K engagements
"Event [--] 🥇: the gold medal task is to wash a frying pan in the sink using soap and water. Both sides. We also tackled silver (cleaning the fingers) and bronze (wiping the counter) in our blog post. The pan is hard to clean but the robot rose to the challenge"
[X Link](https://x.com/physical_int/status/2003161653614125130) 2025-12-22T17:52Z 35.9K followers, 59.3K engagements
"All videos are autonomous. We also tested training "from scratch" (from a VLM initialization) but this failed on all tasks indicating that fine-tuning our models is essential for success. For more check out our blog post: https://www.pi.website/blog/olympics https://www.pi.website/blog/olympics"
[X Link](https://x.com/physical_int/status/2003161655132528917) 2025-12-22T17:52Z 35.9K followers, 41.5K engagements
"To learn more check out our preliminary website:"
[X Link](https://x.com/physical_int/status/1767647869631361227) 2024-03-12T20:24Z [----] followers, [----] engagements
"@Astribot_Inc 🤝 Physical Intelligence ()"
[X Link](https://x.com/physical_int/status/1857441249378410611) 2024-11-15T15:11Z 15.5K followers, 115.8K engagements
"There are great tokenizers for text and images but existing action tokenizers dont work well for dexterous high-frequency control. Were excited to release (and open-source) FAST an efficient tokenizer for robot actions. With FAST we can train dexterous generalist policies via simple next token prediction and get a 5x training speed-up over prior state of the art"
[X Link](https://x.com/physical_int/status/1879963467836453067) 2025-01-16T18:46Z 22K followers, 123.1K engagements
"Compared to prior state-of-the-art VLAs like our own pi0 model FAST policies train 5x faster what used to take weeks can now be trained in days 🦾"
[X Link](https://x.com/physical_int/status/1879963472563417095) 2025-01-16T18:46Z 15.3K followers, [----] engagements
"FAST policies also follow language well and allow us to train the first generalist policies that can perform tasks out of the box in new environments simply by prompting them in natural language"
[X Link](https://x.com/physical_int/status/1879963474287354141) 2025-01-16T18:46Z 15.4K followers, [----] engagements
"Vision-language models can control robots but what if the prompt is too complex for the robot to follow directly We developed a way to get robots to think through complex instructions feedback and interjections. We call it the Hierarchical Interactive Robot (Hi Robot)"
[X Link](https://x.com/physical_int/status/1894829058883731962) 2025-02-26T19:17Z 22.3K followers, 109.8K engagements
"We got a robot to clean up homes that were never seen in its training data Our new model [----] aims to tackle open-world generalization. We took our robot into homes that were not in the training data and asked it to clean kitchens and bedrooms. More below"
[X Link](https://x.com/physical_int/status/1914724966362440148) 2025-04-22T16:56Z 25K followers, 467.4K engagements
"-0.5 performs hierarchical inference inferring high-level semantic subtasks ("pick up the plate") followed by actions. It uses a co-training recipe with data from other robots high-level commands verbal instructions and multimodal data from the web"
[X Link](https://x.com/physical_int/status/1914724969646580006) 2025-04-22T16:56Z 20.3K followers, [----] engagements
"We've added pi-05 to the openpi repo: pi05-base pi05-droid pi05-libero. Also added PyTorch training code🔥 Instructions and code here: This is an updated version of the model we showed cleaning kitchens and bedrooms in April: https://www.pi.website/blog/pi05 https://github.com/Physical-Intelligence/openpi https://www.pi.website/blog/pi05 https://github.com/Physical-Intelligence/openpi"
[X Link](https://x.com/physical_int/status/1965161072413016114) 2025-09-08T21:11Z 35.6K followers, 366.2K engagements
"Our models need to run in real time on real robots but inference with big VLAs takes a long time. We developed Real-Time Action Chunking (RTC) to enable real-time inference with flow matching for the [--] and [---] VLAs More in the thread👇"
[X Link](https://x.com/physical_int/status/1932113398961201245) 2025-06-09T16:31Z 34K followers, 80.7K engagements
"If we use our full pre-trained pi05 model simply finetuning with human video data can double the performance on tasks that are depicted in the human videos"
[X Link](https://x.com/physical_int/status/2001096203749167498) 2025-12-17T01:04Z 33.9K followers, 21.3K engagements
"We were surprised and wanted to understand why. What about [---] enabled emergent human-robot transfer We ran an experiment to test if it only appears above a certain scale. Turns out human transfer scales with the amount & diversity of robot data in VLA pre-training"
[X Link](https://x.com/physical_int/status/2001096205405970804) 2025-12-17T01:04Z 33.9K followers, 60.2K engagements
"Event 1🥇: we fine-tune [---] for the "gold medal" task going through a self-closing lever-handle door. This is hard because the robot has to keep the door open as it goes through it"
[X Link](https://x.com/physical_int/status/2003161640838258832) 2025-12-22T17:52Z 33.9K followers, [----] engagements
"We also did the silver medal task: making a peanut butter sandwich. Very long horizon (open jar spread butter cut the sandwich into elegant triangles and close the jar) lots of force and deformables"
[X Link](https://x.com/physical_int/status/2003161646668394679) 2025-12-22T17:52Z 34.7K followers, 16.1K engagements
"And the bronze: using windex to clean a window in this case one of our phone booths. Just don't get stuck inside while it's cleaning"
[X Link](https://x.com/physical_int/status/2003161648249602356) 2025-12-22T17:52Z 34.7K followers, [----] engagements
"Event [--] 🥈: we tried both gold and silver: peeling an orange and using a dog poop bag. For gold we had to "bend the rules" and use a tool so we believe we only take silver. The dog bag is really hard not least because it blinds the wrist camera when in use:"
[X Link](https://x.com/physical_int/status/2003161650179064119) 2025-12-22T17:52Z 34.1K followers, [----] engagements
"And here is our attempt at the gold. The Olympics requires using the fingers but we had to use a sharper tool. That's a disqualification for us but the result is still really interesting"
[X Link](https://x.com/physical_int/status/2003161651827343749) 2025-12-22T17:52Z 34.1K followers, [----] engagements
"Quantitatively training *0.6 with RL can more than double throughput (number of successful task executions per hour) on the hardest tasks and cut the number of failures by as much as a factor of two"
[X Link](https://x.com/physical_int/status/1990574913422512508) 2025-11-18T00:16Z 35.9K followers, 16.7K engagements
"To learn more see more videos and read a full research paper about Recap and *0.6 see our blog post here: https://pi.website/blog/pistar06 https://pi.website/blog/pistar06"
[X Link](https://x.com/physical_int/status/1990574916039778336) 2025-11-18T00:16Z 35.9K followers, 15.6K engagements
"Check out the blog post and full research paper for more details and experiments including studies into high level vs low level transfer comparisons to robot data and quantifying the utility of wrist cameras. https://www.pi.website/research/human_to_robot https://www.pi.website/research/human_to_robot"
[X Link](https://x.com/physical_int/status/2001096208748863798) 2025-12-17T01:04Z 35.9K followers, 21.6K engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing
@physical_int Physical Intelligence
Physical Intelligence posts on X about in the, robots, more than, gold the most. They currently have [------] followers and [--] posts still getting attention that total [-----] engagements in the last [--] hours.
Engagements: [-----] #
Mentions: [--] #
Followers: [------] #
CreatorRank: [---------] #
Social Influence
Social category influence finance 3.7%
Social topic influence in the 18.52%, robots #883, more than 7.41%, gold 7.41%, silver 7.41%, check 7.41%, paper 7.41%, how to 3.7%, base 3.7%, espresso 3.7%
Top accounts mentioned or mentioned by @anyscalecompute @astribotinc @louround_ @pzrsaa @dwarkesh_sp @trappwurld @robotsdigest @shreyasgite @gchampeau @sophie_defauw @eigenron @humanoidsdaily @chrisbarber @aj_1121 @cryptotrissy @indexventures @jiazhi_yang2024 @wolfejosh @colossusmag @zhiyuan_zhou_
Top assets mentioned Robot Consulting Co., Ltd. (LAWR)
Top Social Posts
Top posts by engagements in the last [--] hours
"We discovered an emergent property of VLAs like 0/0.5/0.6: as we scale up pre-training the model learns to align human videos and robot data This gives us a simple way to leverage human videos. Once [---] knows how to control robots it can naturally learn from human video"
X Link 2025-12-17T01:04Z 35.9K followers, 1.2M engagements
"Our model can now learn from its own experience with RL Our new *0.6 model can more than double throughput over a base model trained without RL and can perform real-world tasks: making espresso drinks folding diverse laundry and assembling boxes. More in the thread below"
X Link 2025-11-18T00:16Z 35.9K followers, 690.4K engagements
"This also shows up in the representations learned by the model. We plot the models representations of human and robot images. As pre-training is scaled up the representation of humans and robots become more aligned: to a scaled-up model human videos "look" like robot demos"
X Link 2025-12-17T01:04Z 35.9K followers, 118K engagements
"We got our robots to wash pans clean windows make peanut butter sandwiches and more Fine-tuning our latest model enables all of these tasks and this has interesting implications for robotics Moravec's paradox and the future of large models in embodied AI. More below"
X Link 2025-12-22T17:52Z 35.9K followers, 536.6K engagements
"Benjie Holson proposed "Robot Olympics" - [--] "events" with gold/silver/bronze medal tasks like washing a pan: These are not tasks we made up ourselves. They illustrate "Moravec's paradox" - everyday tasks we find easy that current robots just can't do. https://generalrobots.substack.com/p/benjies-humanoid-olympic-games https://generalrobots.substack.com/p/benjies-humanoid-olympic-games"
X Link 2025-12-22T17:52Z 35.9K followers, 34.5K engagements
"Event [--] 🥇: the gold medal task is to wash a frying pan in the sink using soap and water. Both sides. We also tackled silver (cleaning the fingers) and bronze (wiping the counter) in our blog post. The pan is hard to clean but the robot rose to the challenge"
X Link 2025-12-22T17:52Z 35.9K followers, 59.3K engagements
"All videos are autonomous. We also tested training "from scratch" (from a VLM initialization) but this failed on all tasks indicating that fine-tuning our models is essential for success. For more check out our blog post: https://www.pi.website/blog/olympics https://www.pi.website/blog/olympics"
X Link 2025-12-22T17:52Z 35.9K followers, 41.5K engagements
"To learn more check out our preliminary website:"
X Link 2024-03-12T20:24Z [----] followers, [----] engagements
"@Astribot_Inc 🤝 Physical Intelligence ()"
X Link 2024-11-15T15:11Z 15.5K followers, 115.8K engagements
"There are great tokenizers for text and images but existing action tokenizers dont work well for dexterous high-frequency control. Were excited to release (and open-source) FAST an efficient tokenizer for robot actions. With FAST we can train dexterous generalist policies via simple next token prediction and get a 5x training speed-up over prior state of the art"
X Link 2025-01-16T18:46Z 22K followers, 123.1K engagements
"Compared to prior state-of-the-art VLAs like our own pi0 model FAST policies train 5x faster what used to take weeks can now be trained in days 🦾"
X Link 2025-01-16T18:46Z 15.3K followers, [----] engagements
"FAST policies also follow language well and allow us to train the first generalist policies that can perform tasks out of the box in new environments simply by prompting them in natural language"
X Link 2025-01-16T18:46Z 15.4K followers, [----] engagements
"Vision-language models can control robots but what if the prompt is too complex for the robot to follow directly We developed a way to get robots to think through complex instructions feedback and interjections. We call it the Hierarchical Interactive Robot (Hi Robot)"
X Link 2025-02-26T19:17Z 22.3K followers, 109.8K engagements
"We got a robot to clean up homes that were never seen in its training data Our new model [----] aims to tackle open-world generalization. We took our robot into homes that were not in the training data and asked it to clean kitchens and bedrooms. More below"
X Link 2025-04-22T16:56Z 25K followers, 467.4K engagements
"-0.5 performs hierarchical inference inferring high-level semantic subtasks ("pick up the plate") followed by actions. It uses a co-training recipe with data from other robots high-level commands verbal instructions and multimodal data from the web"
X Link 2025-04-22T16:56Z 20.3K followers, [----] engagements
"We've added pi-05 to the openpi repo: pi05-base pi05-droid pi05-libero. Also added PyTorch training code🔥 Instructions and code here: This is an updated version of the model we showed cleaning kitchens and bedrooms in April: https://www.pi.website/blog/pi05 https://github.com/Physical-Intelligence/openpi https://www.pi.website/blog/pi05 https://github.com/Physical-Intelligence/openpi"
X Link 2025-09-08T21:11Z 35.6K followers, 366.2K engagements
"Our models need to run in real time on real robots but inference with big VLAs takes a long time. We developed Real-Time Action Chunking (RTC) to enable real-time inference with flow matching for the [--] and [---] VLAs More in the thread👇"
X Link 2025-06-09T16:31Z 34K followers, 80.7K engagements
"If we use our full pre-trained pi05 model simply finetuning with human video data can double the performance on tasks that are depicted in the human videos"
X Link 2025-12-17T01:04Z 33.9K followers, 21.3K engagements
"We were surprised and wanted to understand why. What about [---] enabled emergent human-robot transfer We ran an experiment to test if it only appears above a certain scale. Turns out human transfer scales with the amount & diversity of robot data in VLA pre-training"
X Link 2025-12-17T01:04Z 33.9K followers, 60.2K engagements
"Event 1🥇: we fine-tune [---] for the "gold medal" task going through a self-closing lever-handle door. This is hard because the robot has to keep the door open as it goes through it"
X Link 2025-12-22T17:52Z 33.9K followers, [----] engagements
"We also did the silver medal task: making a peanut butter sandwich. Very long horizon (open jar spread butter cut the sandwich into elegant triangles and close the jar) lots of force and deformables"
X Link 2025-12-22T17:52Z 34.7K followers, 16.1K engagements
"And the bronze: using windex to clean a window in this case one of our phone booths. Just don't get stuck inside while it's cleaning"
X Link 2025-12-22T17:52Z 34.7K followers, [----] engagements
"Event [--] 🥈: we tried both gold and silver: peeling an orange and using a dog poop bag. For gold we had to "bend the rules" and use a tool so we believe we only take silver. The dog bag is really hard not least because it blinds the wrist camera when in use:"
X Link 2025-12-22T17:52Z 34.1K followers, [----] engagements
"And here is our attempt at the gold. The Olympics requires using the fingers but we had to use a sharper tool. That's a disqualification for us but the result is still really interesting"
X Link 2025-12-22T17:52Z 34.1K followers, [----] engagements
"Quantitatively training *0.6 with RL can more than double throughput (number of successful task executions per hour) on the hardest tasks and cut the number of failures by as much as a factor of two"
X Link 2025-11-18T00:16Z 35.9K followers, 16.7K engagements
"To learn more see more videos and read a full research paper about Recap and *0.6 see our blog post here: https://pi.website/blog/pistar06 https://pi.website/blog/pistar06"
X Link 2025-11-18T00:16Z 35.9K followers, 15.6K engagements
"Check out the blog post and full research paper for more details and experiments including studies into high level vs low level transfer comparisons to robot data and quantifying the utility of wrist cameras. https://www.pi.website/research/human_to_robot https://www.pi.website/research/human_to_robot"
X Link 2025-12-17T01:04Z 35.9K followers, 21.6K engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing