Co-design

A University of Michigan team has introduced Auto-Robotist, an LLM-driven co-design agent that proposes structural modifications to robot bodies, trains controllers for each candidate, and validates results in simulation. Its core is a continuously updated skill library that turns each trial into reusable design experience; experiments show improved design search across seven soft-robot tasks in EvoGym. The work was accepted to the EMNLP 2026 main conference and open-sourced.

At its Digital Ecosystem Summit, Tencent Cloud outlined a restructuring of Hunyuan's R&D pipeline: a reorg began in December 2025, post-training infrastructure was upgraded in January 2026 for long-horizon agent reasoning, and major releases now land every two to three months. On evaluation, it supplements public benchmarks with blind tests by 163 internal domain experts and an internal suite spanning nine business lines, addressing the coding-heavy bias of existing agent benchmarks.

OpenAI published a blog post demonstrating the collaborative design capabilities of its latest model, GPT-6 Astra, in physical product development. Employees used spare time to complete a test device design—including enclosure, PCB, and firmware—in under a week, with prototypes manufactured within a month. While showcasing full-stack hardware generation potential, OpenAI acknowledged limitations in handling production constraints and tolerances, requiring human iteration. Additionally, the company announced a research partnership with Mazda for automotive simulation and expanded collaboration with Chip Ganassi Racing to optimize CAD conversion workflows.