# No Memory, No Harness: Why the Database Is the Last Line of Defense — Kay Malcolm, Oracle

## Executive summary

Kay Malcolm argues that while AI models (agents) are powerful, they are incomplete without a robust, centralized memory system. She frames the agent as the 'brain,' the surrounding system as the 'harness' (body), and the database as the 'central nervous system' (memory). The core problem addressed is that current systems (like Git) track code changes, not the human intent or context behind them. To solve this, she advocates for using a unified Oracle AI database to store all five types of agent memory (short-term, long-term, episodic, procedural, and semantic) in a single source of truth, preventing context loss when scaling to enterprise teams.

## Key takeaways

- AI's Limitation: Code vs. Intent: AI tools make individuals faster, but they do not automatically make teams more productive if the context and reasoning behind the code are not shared. Git only records what changed, not the human intent (3:36).
- The Agent Architecture Model: An enterprise agent requires three components: the Model (the brain), the Harness (the body, enabling action), and Memory (the central nervous system, carrying context) (7:14).
- The Five Types of Memory: Effective agent memory must distinguish between: short-term (within a session), long-term (across sessions), episodic (what happened last time), procedural (steps taken), and semantic (meaning) (8:09).
- The Need for a Single Source of Truth: When data is spread across multiple specialized databases (relational, document, graph, vector), agents struggle to reconcile the truth, often guessing incorrectly and wasting tokens (10:54).
- The Solution: Unified Database Memory: A unified database (like the Oracle AI database) is necessary to store all memory types (JSON, relational, graph, vector) in one place, ensuring the agent's memory is non-negotiable and accessible across the entire team (14:24).

## Technical details

- Agent Components: The agent is defined as the Model (the brain), which must be connected to a Harness (the body) and Memory (the central nervous system) to function in an enterprise setting (7:14).
- Memory Types: The five critical memory types are: short-term (session context), long-term (persistent context), episodic (past interactions), procedural (steps taken), and semantic (meaning) (8:09).
- Database Integration: The speaker highlights that a single database can natively store multiple data types—JSON, relational, graph, and vector—allowing for a unified memory store (14:24).
- Implementation: The Oracle agent memory SDK can be used to store live conversations and facts, which are then accessed by the LLM of choice or a local model via the Oracle private AI services container (15:40).

## Practical implications

- When designing complex AI agents for enterprise use, do not treat memory as a simple file system or a single database type; it must be a multi-modal, centralized store.
- Build engineers should prioritize solutions that unify relational, graph, and vector data types to ensure the agent has a single source of truth for context.
- Context sharing must be engineered into the workflow (the 'harness') to ensure that knowledge gained by one team member is available to all others, regardless of time zone or session.

## Topics

AI Agents, Database Architecture, Context Management, Memory Systems, Enterprise AI, Oracle Database, Oracle AI database, PIP install Oracle agent memory, Livelabs.oracle.com

Source: https://www.youtube.com/watch?v=jA_x7F8caHI