Sign InOpen Brain
AI EngineerVideoSource Linked

Intelligence + Continual Learning = Expertise — Yu Su, NeoCognition

Stronger models can still behave like smart novices inside company-specific workflows. Agents need learning loops that compress experience into reusable context, procedures, and judgment without losing reliability.

AI Engineer · Aug 12, 2026
Open Source Open MarkdownOpen JSON
Source Summary

Su separates intelligence—the ability to reason from supplied context—from **expertise**, accumulated competence within a specific environment. Coding agents benefit from a language-native, verifiable world; everyday digital work contains local authority, priority, and exception rules.

Practical Implication

Build learning loops that compress episodes, feedback, facts, and procedures into **reusable structures**. The aim is to help agents select the right context and narrow the search space instead of brute-forcing each task as if it were new.

Agent-Ready Context
Su separates intelligence—the ability to reason from supplied context—from **expertise**, accumulated competence within a specific environment. Coding agents benefit from a language-native, verifiable world; everyday digital work contains local authority, priority, and exception rules.

Build learning loops that compress episodes, feedback, facts, and procedures into **reusable structures**. The aim is to help agents select the right context and narrow the search space instead of brute-forcing each task as if it were new.

Continual learning creates a stability-versus-plasticity problem: agents must absorb new experience without losing dependable behavior. The talk argues that parametric and non-parametric learning are both needed, but does not specify a proven general architecture.
Connected Context · Feed7 Judgment

This reframes agent memory from preserving context to accumulating environment-specific competence. It reinforces external reusable state and episodic-to-parametric consolidation, while adding the stability requirement that learning must not displace dependable behavior. The candidates provide partial storage, handoff, and routing mechanisms, but none resolves that general learning problem.

UniMem: Complementary Episodic-to-Parametric Memory for Boundary-Agnostic Task StreamsUniMem supplies a concrete routing design for the talk’s proposed combination of episodic memory and learned consolidation, while leaving safety and inspectability unresolved.TencentCloud/TencentDB-Agent-MemoryThe memory hub implements the external, reusable side of expertise accumulation through versioned facts, skills, workflows, and code context.Handover of In-Context Learning State Across Session BoundariesTask-relative handover narrows how experience should be compressed across sessions, preserving constraints and irreducible observations rather than generic summaries.AI tools for Forward Deployed Engineering — Vasuman Moza, Varick AgentsVarick’s process-first workflow mapping identifies the local authority, exceptions, and handoffs that an agent must learn before environment-specific expertise is useful.
Context Map
agentcoding#agent-memory#context-engineering#agent-reliability
Uncertainty
Continual learning creates a stability-versus-plasticity problem: agents must absorb new experience without losing dependable behavior. The talk argues that parametric and non-parametric learning are both needed, but does not specify a proven general architecture.