Thailand's first hands-on humanoid school
Phra Khanong, Bangkok
Joined August 2026
- Tweets59
- Following11
- Followers7
- Likes3
So much happened in AI and robotics this week. Standouts from Meta, Fourier, Xiaomi and GPT-6 Astra 🔥 🧵
ALT Humanoice Physical AI Weekly Digest TOP 3 cover: Meta and Fourier Max humanoid, Xiaomi humanoid 8-hour factory task, MIT CSAIL AI edits another AI's weights.
2/ Xiaomi says its humanoid ran an 8-hour task at its EV factory, sorting center-console side covers and correcting anomalies on its own. March: 3 hours. Video reportedly sped up, no throughput data yet.
nitter.gq/XiaomiTech_/status/211…
8 hours on the production line. 🤖
Xiaomi’s full-size humanoid robot takes on an 8-hour task at the Center Console Side-Cover Sorting Workstation at Xiaomi Auto Factory.
Beyond sustained operation, it can autonomously detect and correct anomalies while coordinating with the production line system — another step toward real-world industrial deployment.
#XiaomiRobotics #EmbodiedAI
3/ MIT CSAIL researchers had GPT-6 Astra edit Qwen3.5-0.8B's weights directly: 0.016% changed, no training. MATH-500 7.9% to 48.1% (authors); a rerun got 51.0%. One small model, mostly fixing loops.
nitter.gq/rikiyatakehi/status/21…
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
เหลือแค่ 2 ที่นั่งสุดท้าย ในราคาลดแหกโค้ง 2,900 บาท
คลาส Humanoid Programming 101
(ไม่ต้องมีพื้นฐาน coding)
เราสอนวิธีการควบคุม humanoid โดยเรียนกับหุ่นยนต์จริง!
สมัครได้เลยที่
humanoice.com/th/apply
We put the brain of @GeminiApp ER 2 in @RoboParty_DEV 's humanoid robot. Now it can navigate and reason 🔥
If you own RP0, here is the code we use
github.com/humanoice/gemini-…
So much happened in AI and robotics last week. Here are the standouts from Machenta, Google DeepMind, Tesla and Unitree 🔥🧵
3/ Tesla says Optimus production has started in Fremont 🏭 Musk: 20,000 units a week is at least a year away.
assets-ir.tesla.com/tesla-co…
4/ Unitree's H2-X 🥋 is a full-size humanoid for security and rescue with about 8 hours of endurance. The launch post pulled 15.2M views.
nitter.gq/UnitreeRobotics/status…
ลดด่วน! ลดโหด! 🔥
แค่ 5 วันนี้เท่านั้น
ทุกคอร์สของ Humanoice ลดราคา 50%
เมื่อสมัครช่วง 21 - 26 ตุลาคม 2026
เรียนประกอบ Humanoid เหลือเพียง 4,950 บาท
เรียน humanoid programming เหลือเพียง 1,950 บาท
ถ้าสมัครแพ็คคู่ จะเหลือ 5,520 บาท
ถูกกว่านี้ หาไม่ได้อีกแล้ว
สมัครด่วน
ฝึกควบคุมหุ่นยนต์ด้วย teleoperation
เทคโนโลยีใหม่ ที่ช่วยเก็บข้อมูลเพื่อไปเทรน AI
อยากเรียนรู้เพิ่มเติม ต้องลงคอร์ส humanoid programming 101
เราสอนช่วงวันเสาร์
🕙 10:00-17:00 น.
📍 Humanoice พระโขนง กรุงเทพ
เช็ครอบได้ที่หน้าเว็บไซต์เลย
So much happened in AI and robotics last week. Here are the standouts from Nature Will, Google DeepMind and Physical Intelligence 🔥🧵
2/ Gemini 4 Argon 🎯 watched one cornhole video, mapped it onto Apptronik's Apollo, coded its own MuJoCo sim, practiced, then sank 7 straight throws in a real lab. One general model ran the video-to-sim-to-real loop.
nitter.gq/GoogleDeepMind/status/…
Gemini 4 Argon taught a robot to play cornhole just by watching a video.
It mapped the physics, coded its own simulation, and taught @Apptronik's Apollo humanoid to sink 7 throws in a row.
3/ Physical Intelligence opened the Pi API beta 🔌 Upload robot data, fine-tune Pi's foundation models on managed compute, run inference via a hosted gateway. Robot models are starting to look like LLM APIs.
nitter.gq/physical_int/status/21…
Humanoice retweeted
One of EthPadThai members is launching @humanoicebot a place where you can learn how to build Humanoid Robots in Bangkok.
More info: humanoice.com/
2/ NVIDIA robots assemble GB300 tester trays 🦾 Busbars at 95%+ with a classical stack, connectors at 90-95% with sim RL + real-world force-torque tuning. Lesson: good engineering still beats pure end-to-end.
nitter.gq/NVIDIARobotics/status/…
3/ Meta FAIR's RoboJEPA 📈 shows robot world models follow scaling laws, from 22M to 8B params across 12 robot types. Now you can forecast what compute buys. Checkpoints and code are open.
nitter.gq/artemZholus/status/210…
Introducing RoboJEPA, the first scaling law for multiembodiment world models.
We scaled JEPA predictors up to 8B params, showed scaling laws extrapolate and correlated compute scaling with planning performance.
We release all checkpoints and the code.
robojepa.github.io