Special Edition on Databases

Introduction

Over the past few years, the center of gravity in data systems has begun to shift. While traditional database workloads were dominated by deterministic transactions and analytical queries, both industry and academic evidence now point to a rapid rise in AI-driven, token-based workload. Analysts estimate that 80% of enterprise data is unstructured, yet historically underutilized. Today, systems are being redesigned to make this data queryable through semantic and multimodal interfaces, marking a transition from structured query processing to probabilistic, context-aware data workflows.

In this setting, long-held assumptions are being challenged: LLM invocations increasingly dominate execution cost, latency, and even correctness trade-offs, fundamentally reshaping optimization priorities across the stack. Databases are no longer passive stores but active participants in reasoning loops—supporting retrieval, context assembly, and execution for AI agents.

We’ll look at how practitioners and researchers are rethinking optimization in this new setting, where LLM network calls are the real bottleneck, and efficiency means reducing token usage and managing uncertainty. We’ll also explore how agents are starting to sit between users and databases, turning queries into reasoning loops, how and what data are we vectorizing and what this means for system design. Finally, we’ll touch on open challenges, including working with time-series and evolving data, lack of clear benchmarks and ensuring reliability in probabilistic outputs, and supporting long-running, stateful workflows.

Takeaways

A clear mental model of how data systems are evolving, especially in the semantic query engine space.
Practical insights into what’s changing in systems today, including how teams are optimizing LLM-heavy workloads and translating research ideas into production.

Who is this for?

Engineers and Researchers building or working on databases, data infrastructure, or AI/data platforms