Agents that code the physical world.
Position papers and research from HexaFuture on code as world, code as policy, and coding agents that turn physical feedback into reusable capabilities.
- Articles
- 4 published
- Latest
- November 2026
- Topics
- Physical Coding · Robot tools · VLM agents
Earlier work
Sep–Oct 2026- Research
HexaShell-Preview: A thin harness for exploring the limits of VLM agents as robot coding agents
Our earlier task-tool approach: a thin harness and calibrated RGB views turn robot failures into reusable tools. Official hidden-seed RoboDojo evaluation: 46.85% success and 53.18 progress.
Read article -
Position paper
Self-evolving coding agents: From digital programs to physical-world intelligence
Physical Coding connects code as world and code as policy. We outline how agents can use physical feedback to revise tools, workflows, data, and models, turning verified experience into reusable capabilities.
Read article -
Research
Make code as policy great again: Frontier agents write, call, and evolve robot tools
URAI lets agents write and improve robot tools, while execution models choose actions from fresh observations. Five simulation tasks, seven real-robot tasks, and a tool-evolution study examine this division of work.
Read article
