I have spent more than a decade studying causality. For me, causality is about understanding the rules that govern how actions change the world, and reasoning about their consequences before acting. Scaling has made today's models remarkably capable at spotting patterns and predicting what comes next. A system that acts in the physical world needs more. It has to ask what will happen if it acts, check the outcome, and adapt when reality doesn't match the plan. In the physical world, trial and error can be costly or unsafe, and an early mistake can carry through the rest of a task. When we started Aether AI, robotics was the obvious place to begin. CRIS-0 is our first integrated system built on these ideas. It tracks the conditions that matter for a task, predicts how each action will change them, and replans when something unexpected happens. In our demo, that means taking a metal can off a plate before heating it, and telling an empty can from an unfinished drink while tidying a room. Physical AI is just our first proving ground. We're building towards causal intelligence that also works for forecasting, scientific discovery, and more. The technical write-up is linked in the comments.

Thrilled to share CausalWM, our causal world model v1 built based on Causal Chain-of-Thought (CoT). It currently ranks No. 1 on TriWorldBench and the robot domain of PAI-Bench. Many world models predict the future frames directly from current observations with language or action conditions. This can produce visually convincing videos at first glance, while still getting many subtle but important physical details wrong. Because the underlying physical variables remain entangled in implicit representations, it is difficult to tell whether the model has truly captured the relevant physical mechanisms or is merely relying on shortcuts that produce plausible-looking results. With CausalWM, we bring Causal CoT into future video prediction. The model follows an explicit sequence: Optical Flow → Geometry / Pointmap → Future RGB Each step in the causal CoT captures a meaningful intermediate transition, allowing the model to progressively capture causal dependencies underlying physical evolution. One result I find especially interesting is the in-context control interface created by Causal CoT. Because our CausalWM has learned to use new visual signals in causal CoT. With limited fine-tuning, any useful features (e.g., simulator-rendered robot trajectories) can be added as in-context control, to guide it to generate consistent future videos. This in-context learning ability gives us a practical way for physical reasoning and causal intervention. Links in the thread below.

Can an agent explore a new environment, learn its causal structure, and keep improving without updating its model weights? We introduce RSIAgent, a framework for recursive self-improvement through autonomous exploration. Using Kimi-K3 and GLM-5.3 as base models, RSIAgent outperforms GPT-6 Astra on both OSWorld 2.0 and Agents’ Last Exam. RSIAgent decides what to explore, executes tasks, verifies outcomes, and consolidates stable action-condition-outcome relationships into memory for future use. With the underlying model weights fixed, RSIAgent achieves: - 78.98% Partial Score on OSWorld 2.0 (0808 offline), compared with 72.60% for GPT-6 Astra - 84.82% on Agents’ Last Exam (Near-term), compared with 82.26% for GPT-6 Astra We call this Scaling Experience. Agents can continue improving by acquiring, verifying, and reusing their own experience while the underlying model weights remain fixed. Links in the reply below.