Sight
Back to categories

Explore

Robotics

The strongest cards across the lane. Read the lane first, then drill into the sight behind each card.

AI AGENTS•1 hour ago
3
votes
186
seen

Grok Bot can now control your computer across apps and websites

Grok Bot can now do more than answer questions: it can control a computer, work across apps and websites, and save logins, payment cards and contact information for future use. That pushes it toward an always-on operator for tasks like returns, appointment scheduling, X and Slack monitoring, and opening GitHub pull requests.

Team Bots are available to Teams and Enterprise customers and can run scheduled workflows after a laptop is closed. Version 0.75.2 also adds persistent reports, memos, dashboards and video generation through Krea.

Grok Bot can now do more than answer questions: it can control a computer, work across apps and websites, and save logins, payment cards and contact information for future use. That pushes it toward an always-on operator for tasks like returns, appointment scheduling, X and Slack monitoring, and opening GitHub pull requests.

Team Bots are available to Teams and Enterprise customers and can run scheduled workflows after a laptop is closed. Version 0.75.2 also adds persistent reports, memos, dashboards and video generation through Krea.

yesterday

Grok Bot version 0.75.2 was reported with persistent reports, memos, dashboards and video generation.

11 hours ago

Grok Bot integration with XChat was reported.

5 hours ago

Grok Bot announced that it can save logins, payment cards and contact information for future use.

186

No comments yet. Start the thread.

186
1 comment
View 1 more in Sight
AI FUTURES•3 days ago
136
votes
7.6K
seen

Musk says digital superintelligence could put humans out of control

Musk says digital superintelligence could leave humans unable to control it, even as AI and robots reshape money and government. He argued that safety work should favor curiosity and truth-seeking, with rival systems testing models before release. The economic upside he described is just as extreme: a 20% to 30% jump in global output, or roughly $30 trillion a year.

Musk says digital superintelligence could leave humans unable to control it, even as AI and robots reshape money and government. He argued that safety work should favor curiosity and truth-seeking, with rival systems testing models before release. The economic upside he described is just as extreme: a 20% to 30% jump in global output, or roughly $30 trillion a year.

3 days ago

Musk's interview with Walter Isaacson on The Builders was released, covering AI safety, superintelligence, government, immigration, robots, money and humanity's future.

3 days ago

Musk said digital superintelligence could add about $30 trillion per year to Earth's economy.

7.6K

No comments yet. Start the thread.

7.6K
1 comment
View 1 more in Sight
AI MANUFACTURING•Oct 16
154
votes
6.2K
seen

Taiwan plans a $4 billion Texas complex for AI hardware makers

Taiwan is planning a broader U.S. base for making AI hardware, not just a single factory. The reported Texas complex would cost $4 billion and support Nvidia, TSMC, and other major chipmakers.

That plan comes as TSMC evaluates more Texas capacity beyond its existing U.S. buildout. Its planned investment has reached $265 billion, making the Texas project part of a much larger push to move more of the AI supply chain onto U.S. industrial ground.

Taiwan is planning a broader U.S. base for making AI hardware, not just a single factory. The reported Texas complex would cost $4 billion and support Nvidia, TSMC, and other major chipmakers.

That plan comes as TSMC evaluates more Texas capacity beyond its existing U.S. buildout. Its planned investment has reached $265 billion, making the Texas project part of a much larger push to move more of the AI supply chain onto U.S. industrial ground.

Oct 16

Taiwan's industrial-park plan for a U.S. AI hardware supply-chain base was reported.

Oct 16

The Texas complex was reported as a $4 billion project supporting Nvidia, TSMC, and other major chipmakers.

6.2K

No comments yet. Start the thread.

6.2K
1 comment
View 1 more in Sight
HUMANOID ROBOTS•yesterday
104
votes
4.2K
seen

Tesla’s Optimus must survive beyond a month before outside sales

Tesla is moving Optimus toward outside customers, but the robot still has to clear the basic test: safe, reliable and durable beyond a month of use. Elon Musk said those AI, design and durability issues should be resolved around the end of 2027, when deliveries could begin.

The next proof point is closer. Tesla plans an Optimus V3 video in one or two months showing tasks that are not pre-programmed, while AI5 chip integration is targeted for April 2027. The production goal is 20,000 robots a week, or roughly 1 million a year.

Tesla is moving Optimus toward outside customers, but the robot still has to clear the basic test: safe, reliable and durable beyond a month of use. Elon Musk said those AI, design and durability issues should be resolved around the end of 2027, when deliveries could begin.

The next proof point is closer. Tesla plans an Optimus V3 video in one or two months showing tasks that are not pre-programmed, while AI5 chip integration is targeted for April 2027. The production goal is 20,000 robots a week, or roughly 1 million a year.

2 days ago

Elon Musk said Optimus AI, design and durability issues should be resolved around the end of 2027, when deliveries to outside companies could begin.

2 days ago

Tesla said an Optimus V3 video demonstrating non-pre-programmed tasks could arrive in one or two months, with AI5 chip integration targeted for April 2027.

4.2K

No comments yet. Start the thread.

4.2K
1 comment
View 1 more in Sight
ROBOTICS FUNDING•2 days ago
78
votes
3.1K
seen

Eka Robotics raises $150 million to scale force-sensitive robots

Eka Robotics raised $150 million in a Series A at a reported $1.45 billion valuation, giving the company fresh capital to move its robots into more customer deployments. Andreessen Horowitz and F-Prime Capital led the round.

The company is building a Vision-Force-Action model that combines camera input with force feedback, so a robot can adjust when a grasp misses or contact changes. Eka reported 1,500 picks per hour in lab tasks, putting force-sensitive control against vision-only robot models.

Eka Robotics raised $150 million in a Series A at a reported $1.45 billion valuation, giving the company fresh capital to move its robots into more customer deployments. Andreessen Horowitz and F-Prime Capital led the round.

The company is building a Vision-Force-Action model that combines camera input with force feedback, so a robot can adjust when a grasp misses or contact changes. Eka reported 1,500 picks per hour in lab tasks, putting force-sensitive control against vision-only robot models.

3 days ago

Eka Robotics announced a $150 million Series A led by Andreessen Horowitz and F-Prime Capital.

3 days ago

Eka Robotics was reported to achieve 1,500 picks per hour in lab tasks using vision and force feedback.

3.1K

No comments yet. Start the thread.

3.1K
1 comment
View 1 more in Sight
ROBOTICS AI•Oct 13
172
votes
7K
seen

Natura-1 turns natural-language instructions into dexterous robot actions

Natura-1 turns natural-language instructions into dexterous robot actions without task-specific examples, aiming at the costly bottleneck of hardware-specific teleoperation data. The model handles object placement and orientation, tracks what it has moved or attempted, and chooses the next step during longer tasks while adapting its grip and splitting work between hands. It is a 50-billion-parameter mixture-of-experts model trained on nearly one million hours of egocentric human video.

Natura-1 turns natural-language instructions into dexterous robot actions without task-specific examples, aiming at the costly bottleneck of hardware-specific teleoperation data. The model handles object placement and orientation, tracks what it has moved or attempted, and chooses the next step during longer tasks while adapting its grip and splitting work between hands. It is a 50-billion-parameter mixture-of-experts model trained on nearly one million hours of egocentric human video.

Oct 13

Nature Will introduced Natura-1 as a general foundation model for the physical world.

Oct 13

Additional coverage identified Natura-1 as a 50-billion-parameter mixture-of-experts model trained on one million hours of video.

7K

No comments yet. Start the thread.

7K
1 comment
View 1 more in Sight
HUMANOID ROBOTICS•Oct 12
36
votes
2.4K
seen

Two Booster Robotics K1 robots navigate a live crowd at Wallet

Two physical AI robots operated around a live crowd instead of staying inside a simulation or lab. Axis Robotics brought two Booster Robotics K1 units to the Wallet event on October 7, where they walked through the venue, performed, and interacted with attendees. It was a small public test, but the robots were running in an uncontrolled human setting rather than a video or render.

Two physical AI robots operated around a live crowd instead of staying inside a simulation or lab. Axis Robotics brought two Booster Robotics K1 units to the Wallet event on October 7, where they walked through the venue, performed, and interacted with attendees. It was a small public test, but the robots were running in an uncontrolled human setting rather than a video or render.

2.4K

No comments yet. Start the thread.

2.4K
1 comment
View 1 more in Sight
ROBOTICS MAPPING•6 days ago
3
votes
126
seen

ROS 2 LiDAR localization covered 390 meters, including unmapped sections

ROS 2 mapping tools are moving from isolated demos toward workflows that handle localization, map creation, and mismatched environments. R Sasaki is updating three open-source projects: one localizes 3D LiDAR against point-cloud maps, one builds LiDAR SLAM maps, and visloc-rs brings visual-inertial SLAM and Gaussian Splatting to pure Rust.

The practical test is rougher than a lab loop. A handheld Livox MID-360 covered about 390 meters outdoors on an older map, including stretches outside the mapped area. The latest lidarslam_ros2 demo adds benchmarking and Autoware-compatible map workflows.

ROS 2 mapping tools are moving from isolated demos toward workflows that handle localization, map creation, and mismatched environments. R Sasaki is updating three open-source projects: one localizes 3D LiDAR against point-cloud maps, one builds LiDAR SLAM maps, and visloc-rs brings visual-inertial SLAM and Gaussian Splatting to pure Rust.

The practical test is rougher than a lab loop. A handheld Livox MID-360 covered about 390 meters outdoors on an older map, including stretches outside the mapped area. The latest lidarslam_ros2 demo adds benchmarking and Autoware-compatible map workflows.

Oct 16

lidar_localization_ros2 demonstrated 3D LiDAR localization with a handheld Livox MID-360 across roughly 390 meters outdoors on a prior map.

6 days ago

visloc-rs was highlighted as a pure-Rust project combining structure-from-motion, visual-inertial SLAM, 3D Gaussian Splatting, and localization.

6 days ago

lidarslam_ros2 received an updated hero demonstration covering point-cloud map authoring, benchmarking, and Autoware-compatible workflows.

126

No comments yet. Start the thread.

126
1 comment
View 1 more in Sight
COMPUTE HARDWARE•yesterday
15
votes
1K
seen

RTX Spark laptops get their first hands-on unboxings at GTC Berlin

The RTX Spark laptops are now getting hands-on exposure, not just a launch-stage reveal. NVIDIA GeForce shared community unboxings and reactions from GTC Berlin, including what people are most excited to try on the new laptops. The coverage also included unboxings of GeForce Trading Cards S1 and a cameo from iJustine.

The RTX Spark laptops are now getting hands-on exposure, not just a launch-stage reveal. NVIDIA GeForce shared community unboxings and reactions from GTC Berlin, including what people are most excited to try on the new laptops. The coverage also included unboxings of GeForce Trading Cards S1 and a cameo from iJustine.

yesterday

NVIDIA’s GTC Berlin event highlighted the company’s launch programming and new certification milestones.

yesterday

NVIDIA GeForce shared community reactions and unboxing coverage for the RTX Spark laptop launch.

1K

No comments yet. Start the thread.

1K
1 comment
View 1 more in Sight
ROBOTICS•Oct 12
9
votes
312
seen

UW-trained robots complete dexterous tasks with zero real-world training data

UW-trained robot policies completed gear meshing, nut threading, and peg insertion without any real-world training data. They performed the tasks zero-shot from RGB input, and the same simulation-based recipe transferred to locomotion across difficult terrain.

That moves the bottleneck from collecting demonstrations to making simulations match reality and teaching reliable recovery. Simple, textureless parts transferred well, but a reflective board needed tape to reproduce its appearance, and recovery became unreliable when the board was not bolted down.

UW-trained robot policies completed gear meshing, nut threading, and peg insertion without any real-world training data. They performed the tasks zero-shot from RGB input, and the same simulation-based recipe transferred to locomotion across difficult terrain.

That moves the bottleneck from collecting demonstrations to making simulations match reality and teaching reliable recovery. Simple, textureless parts transferred well, but a reflective board needed tape to reproduce its appearance, and recovery became unreliable when the board was not bolted down.

Oct 11

University of Washington robotics work was highlighted for performing dexterous manipulation without real-world training data.

Oct 12

Agility Robotics emphasized that reliable, repeatable deployment matters more than a one-time demonstration.

312

No comments yet. Start the thread.

312
1 comment
View 1 more in Sight