Fly-by-Code: coding agents pilot drones via active visual and physical feedback loops
jbhuang0604 · x · 2026-11-09
Researchers from SNU, CMU, University of Maryland and others present Fly-by-Code: embodied coding agents for aerial manipulation. The agent writes a task program, executes it in simulation, and — beyond standard execution feedback (console logs, camera images) — writes a separate diagnostic program each round to gather Active Feedback.
- View: when part of the outcome is unseen, the agent flies to new viewpoints, returns images with notes, and uses them to decide whether the task is done and how to revise the program.
- Probe: when failure causes are unclear, it runs targeted physical tests (e.g., a straight pull fails to open a door, an arc motion succeeds) to acquire physical evidence.
The gathered evidence guides iterative task-program refinement, closing the two blind spots of conventional code-and-execution loops. Code coming soon.
More from coding & agent
- Dev rebuilds Pokémon Red in 2.5D from his own ROM cartridge using an agentic workflow — NathanWilbanks_ · 2026-11-10
- Same prompt shootout: Claude vs ChatGPT vs Figma Make on an e-commerce landing page — Tegadesigns · 2026-11-10
- Publishing an MCP server to the official registry: the 403 gotcha and immutable versions — srikanth_builds · 2026-11-10
- Sonner creator ships /gestures skill packing rubber-banding and momentum drags for apps — floguo · 2026-11-10
- NotiBuddy: Open-Source MCP Tool Lets Coding Agents Ping Your iPhone via Lockscreen — dbenc · 2026-11-10
- Skyrim modder uses AI to reverse-engineer the 15-year-old engine for real-time terrain editing — 141_1337 · 2026-11-10