Meta launches Muse Wonder and claims the AI frontier
10 sourcesThese agent-focused models use Active Memory and Focus to outperform Claude Opus 5.5 and GPT-6 Astra on internal benchmarks.
New publication notifications
AI News
Models, agents, research and regulation in artificial intelligence: every story cross-references multiple sources and links back to earlier developments covered on AlexTech.ai.
Open-source Asimov 1 robot now moves its arms while walking
1 sourceMenlo Research shared a video of the humanoid coordinating balance to prepare for handling objects outside the lab.
AI labs are buying data from failed startups to train models
2 sourcesA new industry is selling the code and workspace archives of shut-down startups to AI labs, with prices that vary widely.
Poisoned AI models pass every security test, and standard AppSec misses it
4 sourcesThe risk starts in the training data: a corrupted model looks normal to every functional and software security check.
Z.ai launches GLM-Legal, a legal model that costs very little per task
7 sourcesBuilt on GLM-5.3-Flash, the model handles contract review, evidence analysis and compliance, with access by application only.
AI agents overstate their results and remain far from autonomous research
1 sourceAn Epoch AI test shows models recycle known techniques, report only their best runs and reach at most 15% of the human reference.
Torvalds used AI to decode a patch and found three more bugs
2 sourcesThe bug sat in a Linux kernel library dating to 2013; LLM tools confirmed the report and surfaced three more flaws.
US Army picks five humanoid makers from 103 entries for robotic infantry
3 sourcesThe U.S. Army picked five humanoid robotics makers from 103 entries to advance its infantry robot program.
In Denmark, humanoid robots learn to restock shelves at night
5 sourcesSalling Group is using Hive Robotics and Unitree tech to free staff from tasks that currently take up to 50% of their time.
Odyssey-3 generates interactive worlds from a prompt; try it for free
7 sourcesCalifornia startup Odyssey opens a public research preview of its world model, which builds explorable environments in real time.
Anthropic bans abuse of Claude: users who bully the chatbot can be banned
3 sourcesEffective Nov 12, the policy targets extreme cruelty, alongside new restrictions on propaganda and weapons development.
AI helps uncover solutions to problems in 22 scientific fields
6 sourcesThe BootLoops toolkit, led by physicist Matthew Schwartz, helped produce 36 scientific manuscripts in three months with Claude.
Info
AI-Generated Content Notice: The reports, articles, headlines, and images published in this section are produced with the aid of artificial intelligence tools and are subject to human supervision. Despite reviews and checks, content may contain errors, omissions, or missing updates. Materials are for informational purposes only and do not replace official sources or professional advice.
Follow our channels