Choose an edition
Edition: Enthusiast. How new technologies work and what they change.
Physical AI — the path to a beautiful tomorrow.
THE REPLICANT TRUTHA NEWSPAPER ABOUT PHYSICAL ARTIFICIAL INTELLIGENCE
Forecast · 2026-11-01 · CET
MORNING · 09:00
9.2°
Cloudy
Wind: NW 0.6 m/s · 0%
MIDDAY · 12:00
10.7°
Cloudy
Wind: N 1.4 m/s · 0%
EVENING · 18:00
10°
Cloudy
Wind: NW 1.2 m/s · 4%
Daily: 7.3–11.3 °C
Precipitation probability: 8%
All we can do is learn to be human while the rain falls.
Issue № 4 ·
rev. 3 · Archive
How robots learn to change tools, move and carry out sequences of actions.
Different kitchen actions require different tools: a knife cuts, a brush spreads liquid and a spatula turns food. Researchers led by Yunchu Zhang introduced SwapBot, a system that lets a robot arm exchange tools autonomously. The October 29, 2026 preprint describes a research prototype rather than a household appliance ready to cook without supervision.
The engineering question is how execution works and which constraints were actually tested. A learned initial motion proposal is refined using model-predictive path-integral optimization and learned forward dynamics. A single robot-wagon system, short test routes and a 15% failure rate leave open questions about general deployment.
The engineering question is how execution works and which constraints were actually tested. A diffusion-based alignment policy is combined with a contact classifier and a force-controlled unfastening controller. The initial engagement stage remains unreliable across variations; angled fasteners were not tested.
A humanoid that walks to a table must also stop, bend, reach and recover its balance without switching to an entirely new motor system. Yet many robotics pipelines train separate controllers for each behavior. Chen Tessler and colleagues introduced Metatron, a programmable whole-body execution layer that pairs one trained generative motor policy with software-defined skills. The research was submitted on October 27, 2026; a later listing date should not be mistaken for its original publication date.
Packing an order is not simply a matter of lifting a box. A robot must distinguish small products, position them without damage and handle changes in shape or material. On October 30, 2026, RLWRLD reported a public demonstration of this workflow at KDDI SUMMIT in Tokyo. The event itself took place on October 27 and 28, so the announcement date should not be confused with the demonstration dates. The company described RLDX-1 as its robotics foundation model for controlling physical actions.
Research and demonstrations expand the range of robot actions, but the strength of the result depends on test conditions.
Concepts & terms
The key ideas in this issue, explained briefly.