Software and AI-generated code is moving faster than ever, but speed without safeguards leads to outages, incidents, and costly rollbacks. At Galaxy 2025, we explored how modern teams are transforming the way they release and optimize software in runtime to deliver software and AI applications that are both high-velocity and high-resilience.
We announced innovations in these key areas:
- Guarded Releases: Auto-Generated Metrics, Error Monitoring, Session Replay, and more
- AI Configs: User targeting, experiments, enhanced governance, and more
- Warehouse-Experimentation & Product Analytics: Native Snowflake support, warehouse-native analytics, and more
- Release Management: Launch Insights, Fastly Edge SDK, and several usability and performance updates
Whether you joined us in person or are catching up by watching our recorded sessions, we hope you’re inspired to build
Guarded Releases: Safer rollouts, zero downtime, less manual work
At Galaxy 2025, we announced enhancements to Guarded Releases to help teams monitor their rollouts and stop issues before they become major incidents.
With Smart Minimums, we now dynamically adjust sample sizes based on traffic and release context, reducing false positives and ensuring more confident rollout decisions.
New Health Checks for metrics validate your flag configuration upfront, catching issues before rollout begins. And with Auto-Generated Metrics, teams can instantly monitor feature impact using standardized, out-of-the-box telemetry like errors, performance vitals, and conversion signals.
We also unveiled a powerful new release monitoring view that enables teams to track rollout-specific metrics in real time—and introduced Error Monitoring and Session Replay, now in early access, as part of the Guarded Releases experience.
Error Monitoring detects and groups exceptions triggered by a rollout, surfacing issues with full stack traces, frequency trends, and impacted users—so developers can pinpoint root cause without leaving LaunchDarkly.
Session Replay captures pixel-perfect user sessions linked to feature flags, giving teams a visual timeline of every click, page load, console log, and network request. By combining these tools with flag-level telemetry, teams don’t just spot regressions—they can see exactly how they happened, understand user impact instantly, and debug faster with all the context in one place.
Finally, Guarded Releases is now supported in the EU, giving teams in regulated or privacy-conscious regions the ability to adopt safer, progressive delivery with regional data residency to help meet their compliance needs.
Sign up for Error Monitoring and Session Replay early access.
Watch a quick demo:
AI Configs: Bring great AI apps to market, faster
We unveiled new capabilities for AI Configs—your control plane for managing prompt and model configurations at runtime. The new enhancements give teams even more control over how AI applications are deployed and released, with safety and scale in mind. You can target specific user segments, validate new models in real-world conditions, and run side-by-side experiments to compare performance, all without needing code changes or redeployments.
We also introduced enhanced governance tools that make it easier to review, audit, and manage AI changes. Whether it’s fine-tuning prompts, switching models, or scaling AI across products, the latest updates to AI Configs help teams improve both the quality and consistency of their AI-generated output while also streamlining the development process.
Ready to get started with faster, safer AI? Sign up for our free trial.
Check out a quick demo:
Experimentation: Warehouse-native insights, built into every release
We revealed major updates to LaunchDarkly Experimentation. These updates make it easier for developers to integrate experiments into their workflows and give product teams the trusted insights they need to make better decisions.
Experimentation now connects directly to your data warehouse, with native support for Snowflake. Teams can define and validate metrics using real, production data, no duplication or custom pipelines, ensuring results are accurate, consistent, and aligned with business goals. Developers get an experimentation process that feels like a natural extension of feature flagging. Product managers get clear, reliable insights without chasing down analysts.
We also redesigned the workflow for creating experiments, streamlining collaboration across roles, and removing the guesswork from setup. Previews show exactly how an experiment will run before it’s launched, giving teams confidence from day one.
Finally, we previewed Multi-Armed Bandits, a smarter way to optimize key metrics faster. Instead of waiting for a winner at the end of an A/B test, bandits gradually shift traffic to the best-performing variation in real time—reducing the time customers are exposed to underperforming options.
With LaunchDarkly Experimentation, developers and product managers stay aligned, data stays trustworthy, and every release becomes a chance to learn and improve.
Watch a quick demo:
Product Analytics: Embed powerful analytics into every release
We launched Product Analytics, designed to help product teams go beyond feature delivery and understand how users engage with every release. It enables teams to analyze feature usage, discover patterns, and tie user behavior to business outcomes—all without SQL or waiting on engineering.
With warehouse-native analytics, teams can run queries directly on their Snowflake, BigQuery, or Databricks environments, eliminating data duplication and ensuring a single source of truth. Our self-serve dashboards let users build funnels, explore retention trends, and define behavioral cohorts in just a few clicks.
Product Analytics is also integrated with LaunchDarkly Experimentation. Teams can seamlessly move from insight to action—monitoring feature adoption, running A/B tests, and analyzing experiment results—all within the same platform. You can track how different segments respond to new features, optimize for key KPIs like retention or conversions, and make confident decisions.
See a demo here:
Release Management: Insights and performance upgrades to better manage flags at scale
We launched several updates to help organizations standardize and scale their software delivery practices.
Launch Insights now supports team-level comparisons, so you can see how different groups are using best practices like feature flags, rollouts, and experiments—and where there’s room to improve, like flag cleanup or targeting. This gives you a clear view of release maturity across teams.
To further simplify flag management at scale, Views makes it easier to manage flags at scale. Filter, group, and lock down access based on service, team, or custom tags, so everyone sees only what they need.
We’ve also upgraded performance with FDNv2, our next-gen delivery network, and the Fastly Edge SDK, enabling low-latency targeting at the edge. Advanced users can now use JSON Flag Targeting to define complex rules in the GUI or JSON, with live previews.
Finally, our revamped docs are faster, smarter, and easier to use—with embedded API references and better tutorials to help you build faster.
Check out a demo:
Go forth and build boldly
These releases aim to help teams release faster, mitigate failures instantly, and turn software releases into a competitive advantage. We're excited to see how you'll use these new capabilities across Guarded Releases, AI Configs, Warehouse-native Experimentation, Product Analytics, and Release Management to build boldly, ship fearlessly, and optimize relentlessly.
Like what you read?
Get a demo
