first principles | @imperialcollege alum
Joined July 2021
- Tweets12.3K
- Following1.5K
- Followers43.9K
- Likes48.6K
Are humanoid robots joining workforce?
China’s 2025 humanoid robot shipments:
🔬 Research & education | 42%
🎭 Entertainment & interaction | 35%
🗄️ Data collection | 19%
🏭 Industry & logistics | 4%
Only 4% were shipped for industrial and logistics work.
This highlights two major challenges:
1. Robot intelligence is still too unreliable for many workplace tasks.
2. Deploying robots at scale remains difficult.
Source: New Strategy Research & HRAA, 2025–2026 industry report.
Robotics is entering the API era.
★ Dream Labs | Dream-1 | From $3/hr
★ Google DeepMind | ER 2 | $1 in/$5 out
★ Generalist | Waitlists
★Physical Intelligence | Waitlists
Robot manufacturers can focus on hardware instead of building their own AI models.
Specialized companies can provide robot intelligence through APIs.
The challenge is making these models reliable across different robots and tasks.
End effector types for robots:
🇨🇳 Unitree Dex1-1 | 1 DoF
🇺🇸 Sunday Robotics Memo | 4 DoF
🇰🇷 Tesollo DELTO | 12 DoF
🇸🇬 Sharpa Wave | 22 DoF
Grippers, 3 fingers or 5 fingers...
The best choice depends on the task.
Sunday reports 99.1% laundry-folding success with its 4-DoF hands.
What is 🇨🇳China doing differently?
China is seen as frontrunner in robotics.
As DroidUp's bionic robot plays with children, take a look at the numbers:
An ecosystem at scale:
✸ 243 robotics firm
✸ 161 companies focused on humanoids
✸ ~19K humanoids shipped | H1 2026
That scale is becoming part of everyday life.
Starting in schools:
> UBTECH | Robotics education in 3K+ schools
> 90K+ students in a 2025 robotics & AI contest
Then into shops, events and sports:
→ Galbot | 100+ robot shops across 20+ cities
→ Qingtian Rental | 200 rentals per day
→ WHRG 2026 | 2,056 robots across 666 teams
A generation growing up around robots.
A public getting used to living alongside them.
Deployment is still the biggest challenge.
According to Capgemini, 76% of robotics execs say so.
It's not just technical. Adoption matters too.
One small step:
🇬🇧The UK is backing 8 Robotics Adoption Hubs with $53M.
✸They'll offer advice, studies, trials and training.
✸Companies can test before investing.
Building robots is hard. Getting customers to use them can be just as hard.
Could adoption programs close the gap?
“Our world is designed for the human form.”
Brett Adcock said this when Figure AI emerged from stealth in 2023.
It's one of the biggest reasons behind the humanoid design.
Imagine we've completely solved robotics and AI.
Now imagine the ideal robot is an eight-legged mechanical spider.
Would we rebuild every factory, supermarket, and home to accommodate it?
Even if a better design exists, adapting the entire world to it would be incredibly difficult.
So instead of adapting the world to robots, we're adapting robots to the world.
How much can wheeled humanoids lift?
🇨🇳 S1 | 50 kg | Galbot
🇰🇷 FRIDAY | 20 kg | Holiday Robotics
🇨🇳 Cruzr S2 | 15 kg | UBTECH
🇨🇳 Cruzr Y1 | 15 kg | UBTECH
Payload capacity matters for industrial robots.
Why does this robot feel like an NPC?
Humanoids walk better than ever.
Yet several problems remain:
✸ Many demos still rely on teleoperation.
✸ Robot intelligence remains unsolved.
✸ Whole-body control is still limited.
✸ Skills rarely transfer across robots.
Walking ≠ Intelligence.
Boston Dynamics has a new CEO:
Rohit Prasad.
What does this mean?
At Amazon, he led Alexa AI:
✸ Built and scaled Alexa
✸ Created Amazon Nova models
✸ Led Amazon's AGI efforts
He is an AI-first executive.
Its last two CEOs, Marc Raibert and Robert Playter, were hardware-first leaders.
Boston Dynamics is known for agile robots.
This change at the top may signal a bigger push into AI.
