What question did this study set out to answer?

The main aim is to develop pi-canon as a memory extension for Pi coding agents to improve knowledge recall and management.

August 13, 2026Open Access

pi-canon: Mutable Canonical Memory over an Immutable Journal, with Recall by Surfacing. Design, and a four-arm multi-session evaluation

View Full Paper

Ask AI

Bookmark

Authors

SCShane Conner

Discussion

Member takes

Implication

Randomized trial evaluates recall performance and trap avoidance in a multi-session agent environment, suggesting significant implications for knowledge management.

Key Points

The main aim is to develop pi-canon as a memory extension for Pi coding agents to improve knowledge recall and management.
Conducted a four-arm evaluation with five author-built chains, each repeated five times.
Measured trap avoidance and recall performance across different arm designs under structured protocols.
Utilized an append-only journal for source preservation and implemented budget constraints for recall surfacing.
Canon passed 19 of 20 trap cells, outperforming a bare floor benchmark which passed only 8 of 20 cells.
Plant-only recall achieved scores between 40 to 42 out of 45 across all arms without robust differentiation under the judge.
Canon's recall was at 0.34x the median token count of the bare arm, while the doctrine-file arm consistently performed lowest on session costs.

Cite This Study

Shane Conner (2026) studied this question.

synapsesocial.com/papers/6a7d76bd2b0e0cff3f640458 https://doi.org/10.5281/zenodo.21890648

View Full Paper

Ask AI

Bookmark

Implication

Randomized trial evaluates recall performance and trap avoidance in a multi-session agent environment, suggesting significant implications for knowledge management.

pi-canon: Mutable Canonical Memory over an Immutable Journal, with Recall by Surfacing. Design, and a four-arm multi-session evaluation

Discussion

Member takes

Implication

Key Points

Cite This Study

pi-canon: Mutable Canonical Memory over an Immutable Journal, with Recall by Surfacing. Design, and a four-arm multi-session evaluation

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider