September 13, 2026
Quick hits from across the AI world in the last 24 hours.
- ⚡ Quick hits
- Industry Microsoft said it will integrate xAI’s Grok models into Word, Excel, and PowerPoint, giving Office users an alternative to OpenAI models. IT之家 ↗
- Industry Zurich Airport said it is deploying two WeRide L4 autonomous electric shuttles initially to transport airport employees. Zurich Airport ↗
- Research Fable 5.1 reportedly solved the 370-year-old Cyphral Distich cipher in an automated cryptanalysis demonstration. Vals AI ↗
- Research Researchers described a hierarchical NeRF system using JAX3D for volumetric rendering, novel-view synthesis, and 3D reconstruction. MarkTechPost ↗
- AIGC Figma announced a new episode featuring designer and creative technologist Aitana Picó in its Source Material series. X @figma ↗
- Industry GitHub posted a programmer-focused update encouraging developers to use the platform for coding work and collaboration. X @github ↗
- Research A Princeton researcher proposed a recurrent looped transformer that carries decoder state across tokens instead of using a fixed stack of blocks. MarkTechPost ↗
- Models A Princeton researcher proposed recurrent looping as a way to provide effectively unbounded temporal depth while reusing decoder parameters. MarkTechPost ↗
- Agents Supio said its long-horizon legal agents can execute complex case workflows across systems and communication channels. SiliconANGLE ↗
- AIGC Cohere promoted an AI-art performance titled Convergence at Berlin Art Week featuring musicians Actress and Sega Bodega. X @cohere ↗
- Safety Astra and Fable were reported to remain vulnerable to simple variants of alignment evaluations introduced in 2025. LessWrong ↗
- Open Source A GitHub project called CUDA for AMD on Windows appeared as a community effort to make CUDA-oriented workloads more accessible on AMD hardware. GitHub ↗
- Agents Databricks promoted an SSH tunnel that lets developers use VS Code, Cursor, or terminals while running code on Databricks compute. X @databricks ↗
- AIGC ElevenLabs made its Music v2.5 model available through its app and API with free and paid tiers. The Decoder ↗
- AIGC Lovart showcased GPT-Image 2.5 arranging flowers from a natural-language instruction. X @lovart_ai ↗