Verified Events
DateTypeCompanyHeadlineSegment
Physical Intelligence opens Pi API beta to more robot developers
- Teams can upload their robot data, launch fine-tuning on PI-managed compute and infrastructure, and run the resulting models through hosted inference.
- Early users include Argon Robotics, which trained a dish-racking policy within 35 hours, and General Instinct, which trained a server-assembly policy within three days.
- Beta participation is by application; most showcased applications use cloud-hosted inference, with network connectivity identified as an area for improvement.
Odyssey releases Odyssey-3 foundation world model research preview
- Odyssey reports Pro scores of 66.1 on Physics-IQ Verified video-to-video and 54.7 on image-to-video using prompt enhancement and best-of-eight selection.
- The company reports first place in three of four WorldMark categories in its own evaluations.
- Robot demonstrations include recovery from missed grasps and dropped objects; physical-system adaptation uses learned action decoders or policies.
Aether AI demonstrates CRIS-0 causal robotic intelligence system
- The agent decomposes a small set of demonstrations into verifiable stages and selects rule-based functions, learned policies, navigation and world-model tools.
- When verification fails, the system can re-grasp, revisit a prior stage, replan or revise a failing tool implementation.
- Future-video samples generated by the world model are presented separately from the physical robot demonstrations.
TwelveLabs releases Pegasus 1.6 for video understanding
- Documented uses include action labeling, quality checks of Physical-AI footage and prediction of an operator’s next action.
- The model supports a 261,120-token context window and analysis of one to twenty images per request.
- Video segmentation can return timestamped events within segments, alongside improved entity recognition and metadata extraction.
Field AI reportedly signs $700M financing term sheet at $10B valuation
- The proposed valuation would be five times the company’s previous $2 billion valuation.
- Combined revenue and customer contracts total more than $135 million, including at least $35 million added since June.
- The earlier commercial milestone covered more than 30 customers spanning construction, data centres and defence.
Field AI and General Robotics validate autonomy with HTX through Dimension X
- Field AI-powered quadrupeds located spill sources and potential casualties without human pre-mapping.
- A separate General Robotics evaluation used GRID for elevator-button pressing and corridor navigation.
- The work involved HTX’s Robotics, Automation and Unmanned Systems Centre of Expertise through Dimension X Cohort 5.
Runway introduces Praxis-1 world action model for real robots
- Noble Machines, Standard Bots and Ultra are testing the model on their hardware, including the Standard Bots RO1 arm.
- The model learns robot actions from large-scale video pretraining combined with robot interaction data.
- Public weights are planned in the coming months, following the pre-launch partner programme.
General Intuition raises $220M at a $6.2B valuation
- Valor Equity Partners led the financing, with Atreides, Seven Seven Six, Point72, Khosla Ventures and General Catalyst participating.
- The capital supports expansion of its research and engineering teams in New York and Europe.
- The company opened a waitlist for partners seeking early model access across games, simulation and robotics.
Skild AI publishes physical self-play result with humanoid transfer
- The S1-based policy was rewarded for scoring rather than for hand-designed dribbling or tackling behaviours.
- Dribbling, shielding, tackling and recovery emerged as the policy competed against successive versions of itself in NVIDIA Isaac Sim.
- Skild says it is extending the same post-training approach to everyday robot tasks beyond sport.
Aether AI introduces CausalWM embodied causal world model
- Aether reports a 66.04 TWB score on TriWorldBench and top robot-domain performance on PAI-Bench.
- The training approach combines video pretraining, causal chain-of-thought mid-training and reinforcement-learning post-training.