Awesome Code-as-X: curated papers unifying agent research via programs as representations
ccloy · x · 2026-09-18
A new curated GitHub repo, awesome-code-as-x (65 stars), collects coding-agent papers that treat a program as an intermediate representation across four directions:
- Code as Policy: LLM/VLM writing executable robot control programs
- Code as World: programs representing explicit scene/world state executed by simulators or renderers
- Programmable World Models: programmable world models
- Edit Graphs: program/graph structures controlling image and video editing
It also includes related benchmarks. The unifying idea: programs as executable, inspectable, revisable representations. Continuously updated, PRs welcome.
More from coding & agent
- Open-source repo collects the best JEV use cases in one place, PRs welcome — matchaman11 · 2026-09-18
- User: Astra in Codex is unusable even on a 20x subscription — OpenAI should copy Anthropic's usage-quota approach — CtrlAltDwayne · 2026-09-18
- Controlling the iOS simulator with Jev + AXe: ultra fast at a fraction of LLM cost — stefanjblos · 2026-09-18
- Jev reviews PRs at $0.00007 each, ~200x cheaper than Claude Opus with 14 typed checks — stefanjblos · 2026-09-18
- Developer who abandoned notebooks for agents says Codex made them fun again — intellectronica · 2026-09-18
- ScrapeGraphAI: open-source library that scrapes any site with one plain-English prompt — thisdudelikesAI · 2026-09-18