@m_wulfmeier
Mastery-first Physical AI @NomagicAI, Affiliate @mila_quebec Prev @GoogleDeepMind, PhD @UniOfOxford, visiting @UCBerkeley @ETH @MIT, European @ELLISforEurope
AI
CoRL 2026 accepted 687 papers from a record 2,094 submissions, a program chair announced.
AI
A post says Carbon scored 88.2% on SEC-bench Pro and is rolling out first to cybersecurity specialists.
AI
In an Oct. 22 post, @m_wulfmeier outlined plans to co-advise students and collaborate on physical AI research.
AI
A post calls it one of FAR's most surprising recent projects and links to the collaborators' work.
XAI
A post calls for simpler RL and welcomes what its author sees as momentum for critic-only approaches.
AI
The author says their write-up details how simulation complements real deployment in mastery-first physical AI.
AI
A post says one commit references a K3.1 deployment already running on Deepinfra.
AI
A poster sharing Mila's notice asks applicants interested in working with them to mention their name in the application.
AI
An MLinPL attendee says 800 engineers and researchers gathered, and argues physical AI is more nuanced than social-media portrayals.
AI
The speaker praised Warsaw’s tech talent and invited people to ask where they’re hiring before the talk.
AI
An attendee thanked ETH AI Center for the invitation and said the conference brought “many surprises.”
AI
An attendee representing the NFAI jury at SPRIND’s AI & Robotics Summit in Berlin says the challenge teams make them hopeful about robotics and future SPRIND challenges.
AI
One post predicts that big models will soon challenge the idea that they’re too slow for robotics. A quote-post argues that most such systems need little on-the-fly reasoning for real-time behavior.
AI
A contributor to “Visual Storytelling: An Embodied Companion” calls expert feedback on the drawing robots eye-opening, emphasizing the value of learning through deployment.
AI
LeRobot says the dataset spans 1,274 hours recorded in 27 homes and is available under a CC-BY-4.0 license.
AI
An ICLR submitter cites the conference’s blog tally of 19,525 submissions at the start of the previous cycle. A reply raises concerns that faster AI-assisted research encourages quantity over depth.
AI
Figure says its robots began useful work in 30 rented Bay Area homes without additional training. A team member reports 56% zero-shot success with Index pretraining, versus 8% without it.
AI
A speaker says they will discuss physical AI and the importance of deployment at the summit on October 1, 2026.
AI
A user hopes for robotics training that needs less data, but doubts that a better training objective will look very different from next-token prediction.
AI
A post credits the DARPA Robotics Challenge with enabling progress and floats another contest, arguing that handling objects remains hard and specialized setups often win.
