Agents

The repository implements the classic Monte Carlo AIXI approximation path and exposes it from the same infotheory aixi <config.json> entrypoint used by the rest of the agent stack [veness2011_mcaixi]. In the current codebase, MC-AIXI is the default planner path behind planner: "mc-aixi".

What matters structurally is that the planner is not tied to one single world model. The implementation is built around a shared predictor interface, so CTW/FAC-CTW remain the canonical paper-aligned choices, while other backends can still be plugged in where appropriate.

AIQI is present as a first-class planner path, not a note for future work. The implementation follows the model-free H-step return framing from the AIQI paper and validates the paper-critical constraints directly in configuration: N >= H, discount_gamma in (0, 1), and baseline_exploration in (0, 1] [kim2026_aiqi].

The paper-faithful path is algorithm: "ac-ctw" or algorithm: "ctw". The codebase also extends AIQI to other compatible predictors, while intentionally rejecting ZPAQ-backed strict conditioning.

Both planners sit on top of the same environment abstraction. That shared layer matters for benchmarking because it means the same built-in environments can be used while swapping only the planner.

Today the repository includes built-in environments such as coin-flip, Kuhn poker, extended tiger, tic-tac-toe, and biased rock-paper-scissors, with optional Nyx VM-backed environments when the vm feature is enabled.

The Python extension crate exposes both agent paths too. That matters for reproducible experiments, quick scripting, and external baseline comparison against older Python implementations such as PyAIXI.

For browser-based experimentation on the currently exposed surface, see Workspace.