Elastic

Everyone's racing to add AI. The teams pulling ahead aren't winning on models, they're winning on context. At #MSBuild, Elastic is showing developers what that advantage looks like on Azure, using Elasticsearch, Agent Builder, Workflows, and Azure AI to build AI that performs in the real world, not just in demos. Join us June 2–3 in San Francisco: → Live demos you can actually learn from → Technical sessions on context engineering and AI workflows → Real conversations with Elastic engineers and experts Stop by booth F106. See you there. https://go.es.io/4dUGBN6

CPU-hungry vector indexing is a hidden tax on every AI search app at scale. Corey Nolet from NVIDIA and Blake Holden from Elastic are doing a live session on offloading that work to a GPU via Elasticsearch's cuVS plugin, same index, same queries, faster ingest, freed-up CPU. Same Elasticsearch 9.4. One config flag. Register: https://lnkd.in/e4bar8ec

Data is the fuel for AI, but retrieval is the engine. Elastic, Dell Technologies, and NVIDIA are tackling the hardest part of enterprise AI - pulling the right context, fast, from petabytes of unstructured data. Swipe through for perspectives from leaders across all three companies on how retrieval, accelerated infrastructure, and AI are reshaping what enterprise AI looks like at production scale.

Building a time series in Kibana Dashboards used to take 14 clicks. Now it takes three. • Brand new API • Natural language dashboard creation • Plus much much more https://go.es.io/49cExxr

OTel is the fastest-growing project in the CNCF ecosystem after Kubernetes. Here's everything your team needs to know to get started.

Most security teams still rely on spreadsheets, manual checks, and gut instinct to know if their detections are working and their data is reliable. SIEM Readiness changes that. Now in technical preview in Elastic Security 9.4, it helps teams quickly spot gaps in data: ✔ Coverage ✔ Quality ✔ Continuity ✔ Retention With SIEM Readiness, you can catch issues before they become blind spots, not during an investigation or audit. No spreadsheets. No guesswork. Find out if your SIEM is actually ready: https://go.es.io/3RO9p16

Our BBQ at 1-bit/doc beats TurboQuant at 4-bit/doc on shifted data on ranking accuracy. At 1/5 the storage. We center on the segment centroid before quantization, so the bits go where they are actually needed for ranking. TurboQuant's Hadamard rotation can't exploit that structure. The throughput gap is bigger than the accuracy gap. Our symmetric kernels: 7-22 ns/doc. TurboQuant: 216-275 ns/doc. 10-40x on the same CPU. The reason is architectural. Uniform grids decompose into integer dot products — popcount and multiply-accumulate on NEON. Non-uniform centroids force a data-dependent gather for every coordinate pair. The FMA is cheap. The gather is the bottleneck. TurboQuant does win on reconstruction MSE. But most of that advantage is the Hadamard rotation, not the Lloyd-Max centroids. We applied the same rotation to OSQ and the gap nearly vanishes — 0.306 vs 0.307 at 1-bit. More importantly: MSE measures reconstruction quality. Search engines rank by dot products. Those are different objectives, and the gap between them is where our data-dependent design pays off. Full benchmark methodology, reproduction code, and the math behind centroid centering vs. Hadamard rotation: https://go.es.io/3Ru77Eg

New research on software engineering skills found 4 out of 5 do nothing. That's not a rounding error, that's the default outcome. And the ones that hurt performance aren't obviously bad skills. They just fill the context window with instructions the model didn't know how to use. The finding that matters most: a small model with the right skill beats a frontier model without one. JP Hwang breaks it down → https://go.es.io/4dwhEWV

For financial services organizations moving from AI experimentation to implementation, the biggest barrier to scaling agentic AI is not the model itself, it’s the data foundation underneath it. Financial services companies that can readily sift through both their structured and unstructured data, keep it secure, and apply it in the right context will get the most value from agentic AI.

Security teams shouldn’t have to waste hours moving between disconnected tools just to investigate a single alert. Elastic Workflows brings automation directly into Elastic Security, where AI agents can reason through investigations, determine which tools to use, and execute actions in context. That means analysts spend less time stitching together workflows and more time stopping real threats. No separate automation layer. No added operational drag. Just faster, more intelligent security operations built directly on your security data. Watch Elastic's James Spiteri and Tinsae E. show how agents dynamically choose tools and workflows in real time to accelerate investigations and response: https://go.es.io/4tvcZuh