Robotics papers decoded, machines priced, and the chip-industry job market counted — translated from our Korean field lab.
Semiconductors earned 99.7% of Samsung’s operating profit while the device division printed its first loss — yet the phone unit out-paid chips on the regular bonus. The numbers behind the gap, with the fine print.
One equipment maker lost two-thirds of its operating profit and still raised average pay 29%. What the filings, the global margins, and a strange search-result mirage reveal about how the chip-equipment industry really works.
The viral five-year break-even misses opportunity cost and resale recovery — and most recent Teslas in Korea can’t receive the feature at all. The corrected math, step by step, with the VIN check that decides everything.
The real entry price of a four-legged robot, the battery cost nobody advertises, an 11-day rental break-even, and why the most-searched robot dog of all cannot actually be bought.
A walking humanoid you can pay for with a credit card is real now. The full price math on the Unitree R1 — retail versus import with every tax — plus the two traps: no grasping hands, no custom code.
At one trade show, SK Hynix stretched flash sideways (HBF: up to 512 GB, 3.0 TB/s) while Samsung stacked DRAM on top of the accelerator (zHBM). What the two terms mean, how mature each really is, and why the road ends on a factory floor.
Not yet. As of September 28, 2026, Tesla has not unveiled Optimus V3, published a price or a spec sheet, or announced the start of production. Musk’s target for public sales is the end of 2027.
The five ideas that unlock every robotics paper of 2026 — world models, VLAs, simulation, sim2real, and the data problem — explained with no math and linked to real decoded papers.
KAIST’s HOUND took the Science Robotics cover by fusing walking, running, jumping, and ledge-climbing into a single Transformer controller — with 15.5 hours of training data generated in 8 minutes.
TouchWorld imagines the pressure it is about to feel and treats that feeling as the goal — beating Pi-0.5 and GR00T N1.7 on contact-heavy tasks, and degrading least when humans interfere.