Latentic AI explores harness engineering for AI agents: the systems around models that determine how agents act, use tools, retain state, acquire capabilities, and interact with software.
The work is applied to AI in education, for teachers and learners, as well as general-purpose agent systems. Reusable infrastructure and components are published independently.
| Project |
|
| agent-harness |
Drive Claude Code, Codex, OpenCode, and OpenAI-compatible agents from Rust through a unified event stream. |
| viva |
Persistent, self-extending agents in Common Lisp. |
| Project |
|
| Compose |
Local-first AI Markdown editor for macOS. |
| Syntactic |
Learn Rust offline on an iPhone with an on-device WebAssembly compiler. |
