Total Recall: Agent Memory and Harness Engineering — Ignacio Martinez, Oracle
AI Engineer World's Fair 2026 · 1:00:47
← All organizations
Enterprise databases, cloud infrastructure, and applications
Oracle
Oracle builds enterprise databases, cloud infrastructure, and business applications. Oracle Cloud Infrastructure (OCI) supplies computing capacity for AI training and inference, while Fusion Applications supports finance, supply chains, human resources, and customer operations. NetSuite combines accounting, commerce, and resource planning; Oracle Health provides healthcare applications. Its AI Data Platform connects enterprise data with business context so teams can build agents that automate work across Oracle and third-party systems.
Larry Ellison, Bob Miner, and Ed Oates founded the company as Software Development Laboratories in 1977. Today, Clay Magouyrk and Mike Sicilia serve as CEOs, with Ellison as chairman and chief technology officer. Its Oracle AI Database integrates semantic vector search with business-data queries in SQL, allowing developers to search embeddings alongside relational, JSON, and graph data. JSON-Relational Duality exposes relationally stored data as updatable JSON documents, letting applications use either representation without maintaining separate copies.
Oracle’s deployment options span its own cloud, Amazon, Google, and Microsoft cloud data centers, and customer premises through Cloud@Customer. This gives organizations choices about where their databases and applications run, including infrastructure inside their own data centers. Oracle reported $67.4 billion in fiscal 2026 revenue, including $34.0 billion from cloud infrastructure and cloud applications; its business also includes software, services, and hardware.
Explore the recordings
Oracle’s supplied archive currently contains one recording. Use it to explore team coordination, agent memory, and database-backed infrastructure. The guide below describes claims made in the recording; it does not verify today’s product capabilities or availability.
Start with the team coordination problem
In No Memory, No Harness: Why the Database Is the Last Line of Defense, Kay Malcolm describes developers becoming individually faster with AI while testing, validation, and cross-team coordination remain bottlenecks. Follow this path for her account of code moving between teams without its originating agent context, and her argument that shared memory must preserve intent alongside code changes.
Explore memory types and the proposed database foundation
Return to the same recording for Malcolm’s distinctions among short-term, long-term, episodic, procedural, and semantic memory. She then argues against fragmenting memory across separate database systems and presents Oracle AI Database 26ai and the Oracle Agent Memory SDK as a shared-memory approach. Her example connects context to forks, branches, and commits using Oracle Autonomous Database. Treat the capabilities and productivity benefits described here as recorded presentations and claims, rather than confirmed current product facts.
2 talks
Newest first
2 speakers at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.