From Africa/Climate & energy tech/Nigeria · South Africa
A Random Forest model turned ordinary weather records into hourly sunlight forecasts for one site in Ibadan, Nigeria. Its yearly error on total sunlight was 0.19.
Research Radar/AI agents & MCP/USA
ChainWatch would spot attacks that spread across many AI agent tool calls. The authors say plainly it is a design specification that is not running, checked only against five paper scenarios.
Research Radar/Robotics/USA
A humanoid picked loosened screws from a real Hyundai Ioniq 5 battery pack 86.7 percent of the time while standing still. After it walked itself into place, that fell to 36.7 percent.
Research Radar/Mixed reality/Canada
On the same Meta Quest 3, virtual reality users reached 0.31 cm short while camera passthrough users reached 0.43 cm too far. The virtual reality shift also lasted after the headset came off.
From Africa/Agritech/Tanzania
A Tanzanian team trained a Vision Transformer on 100,000 bean leaf images, adding tiny nudges made to fool it. Accuracy rose from 97.4 percent to 99.4 percent.
Research Radar/Robotics/USA · China
TANGO sets all 29 joints of a humanoid at once, so it can duck, squat and sidestep while it walks. Trained only in simulation, it finished 10 of 15 real cluttered runs against 6 for the rival.
From Africa/Biotech & health tech/Côte d'Ivoire · USA · Canada
A team in Cote d'Ivoire relabelled a day of microscope images, retrained the model overnight, and tested it the next morning. Its catch rate rose from 75.5 to 98.1 percent.
Research Radar/5G & 6G/India
A survey of tiny models for 6G reports one study that turned a 4.55 megabyte neural radio protocol into a 1 kilobyte rule set, a 99.98 percent cut with no loss in task performance.
Research Radar/Quantum computing/USA
A review of neutral atom quantum computing reports arrays of 12,001 trap sites and one atom holding its bit for 119 seconds. It is a survey of other labs, not a new experiment.
Research Radar/Robotics/Singapore · China
A single model let a humanoid walk and work at once across 11 household chores, finishing 81.8 percent of runs. The best of nine rival models finished 44.5 percent.
Research Radar/Space tech/United Kingdom
A model of 16 real satellite fleets found nine of them exceed the size at which their own wreckage keeps growing. Six of those nine cross the worst line, the runaway one.
Research Radar/Cybersecurity/China
Eight MCP security scanners flagged 96.89 percent of running servers as risky. A hand check of 100 of those servers found that only 45.53 percent of the alerts were real.