Zero
An agent-collaborative programming language experiment centered on program graphs
About Zero
Zero is an experimental programming language for human-agent collaboration that uses graphs as program representation. Agents can query program graphs, submit checked edits, and validate results; the compiler is also part of the editing loop, making changes checked graph modifications. Graph patches are constrained by semantic nodes, fields, graph hashes, and expected values, so stale or invalid edits fail before being written to storage. The project aims to remain small, fast, explicit, and dependency-free; the official guidance also warns that breaking changes may still occur and recommends running it in a safe environment.
Best for
Best use cases
Let agents query the program graph
Submit validated graph edits
Validate graph changes in a safe environment
Features
Program graph
Uses graphs as program representation; agents can query the program graph.
Checked editing
Graph patches target semantic nodes and fields, guarded by graph hashes and expected values.
Compiler in the loop
Puts the compiler in the agent editing loop so changes are checked first.
Lightweight runtime goal
Aims for a small, fast, explicit, and dependency-free runtime.
Pros
- Edits are checked with the compiler in the loop
- Graph hashes and expected values prevent stale or invalid edits
- Designed to be small, fast, explicit, and dependency-free
Cons
- Still subject to potentially breaking changes
- The official guidance recommends running it in a safe environment and avoiding direct connections to production systems