Building at the frontier of agentic AI

SpaceX IPO'd on Nasdaq under ticker SPCX at $135/share, raising $75 billion — the largest IPO in history. The kicker: xAI is bundled inside, making SPCX a bet on rockets, Starlink, and frontier AI under one roof. First-day close: ~$161, implied market cap ~$2 trillion. The AI IPO era has a new benchmark.

This paper addresses the limitation of large language and vision-language models in maintaining spatial consistency during fine-grained visual editing by introducing a structured reasoning framework that operates over scene graphs. By reformulating text-conditioned spatial editing as explicit graph reasoning rather than end-to-end generation, the method enables precise manipulation of object layouts through natural language instructions while preserving geometric coherence. The work establishes structured scene-graph reasoning as a necessary intermediate representation for bridging high-level linguistic commands with geometrically consistent spatial editing in 3D environments.

Decomposes video world models into two pillars: state construction (implicit — context in transformer activations, vs explicit — compressed latent tokens or 3D scene statistics) and dynamics modeling (how models integrate physical knowledge and architectural priors). Argues evaluation must shift from visual fidelity to functional benchmarks: physical persistence, causal reasoning, and task-supportable dynamics. Identifies two frontier challenges — persistence and causality.