Memory and skills share one system for better learning and reuse
Memory-Skill Isomorphism: One Skill Carrier, Two Native Uses
Software Engineering
Summary
The paper finds that memory and skills, which help improve intelligent agents, can be combined into a single system instead of being managed separately. It shows how using skills to carry both immediate cues and detailed memories keeps the system efficient and scalable. Testing one implementation suggests this method can reduce failure rates on selected tasks. However, results are early and based on limited data, so more work is needed to confirm the benefits.
What this means in practice
- •For machine learning engineers: Design learning agents that store memories and skills together for efficient knowledge reuse and reflection during task performance.
- •For ai system architects: Implement shared storage and retrieval mechanisms for skills and memories to simplify updates and controlled knowledge management.
Authors
Kang Ruiyuan
Abstract
Memory and skills improve agents without changing weights: memory carries prior experience, skills carry reusable procedures. Wrapping both in stores, routers, retrieval, reflection, and update paths makes reuse machinery grow with accumulation. Part of this duplication need not be rebuilt: a Skill is already a natural carrier for distilled memory. Here the memory component is a Skill: resident description holds hot cues, on-demand SKILL.md a colder index and curation policy, and reference/*.md files detailed memories (levels L0 -> L1 -> L2). A governed 1,024-character description budget keeps the resident index compressed. Memory and capability thus share one progressive-disclosure carrier; reflection evolves either Skill. Writes fork: history appended, current state rewritten and revalidated. In one deployed system, the sharpest identification is governance: 4 write entries exist, but only 1/4 reaches the settlement ledger. At the priced operating point, one L1 lesson-1 point adds 1,313 first-turn tokens against a 1,462-token baseline--tied at k=1 (1,313 versus 1,365), 4,949 at k=5; session totals differ by only 1.18x with overlapping last-turn ranges; price decomposition is unavailable. On one selected task with one model, exposing the lesson means fewer failures (0/8 or 1/8 with the lesson versus a shared non-concurrent 6/6 historical floor, unadjusted for multiplicity). The task was selected on prior floor evidence, so this is a selected-task post-selection existence signal, not a confirmatory rate. Resident and BM25 show no detected difference in two small comparisons. This motivates a candidate RSI design rule: one Skill carrier family, a shared read side, governed distinct writes. Body delivery h := P(D|A) is uninstrumented in RQ1--RQ2; this evaluates a resident-index implementation.