Platform Overview Demo

5:16

Platform Overview Demo

LaunchDarkly 07.04.2026 165 просмотров 4 лайков

Machine-readable: Markdown · JSON API · Site index

Смотреть на YouTube

Поделиться Telegram VK Бот

Транскрипт Скачать .md

Анализ с AI

Описание видео

In this walkthrough, Tom breaks down how LaunchDarkly gives you full control over software releases—from traditional features to AI-powered systems. Modern teams ship faster than ever. But speed without control creates risk. LaunchDarkly lets you release safely at runtime, monitor impact in real time, and automatically respond when things go wrong. No redeploys. No guesswork. Just a tight feedback loop between what you ship and how it performs. You’ll learn how to: Release features globally in milliseconds. Roll back instantly when something breaks. Gradually roll out changes with precision targeting. Automatically detect regressions and trigger rollbacks. Monitor release health alongside system health. Manage AI systems with real-time evaluation and guardrails. Turn every feature into an experiment with fast feedback. This is how modern teams move fast without losing control. Chapters 00:00 Introduction and release strategy 00:16 Release at your speed with runtime control 00:35 Instant rollbacks and safe release practices 00:58 Progressive rollouts and automated safeguards 01:20 Observability and self-healing systems 01:43 Managing AI systems in production 02:05 Real-time regression detection and kill switches 02:24 Unified telemetry and end-to-end visibility 02:46 Release-level observability and signal detection 03:08 Moving beyond firefighting to innovation 03:27 Experimentation without bottlenecks 03:50 Turning every feature into an experiment 04:07 Built-in AI for insights and decision support 04:26 Closing the loop between insight and action 05:02 Final thoughts and next steps

Оглавление (15 сегментов)

Introduction and release strategy

Hi, my name is Tom and over the next few minutes I'll give you a high level overview of LaunchDarkly end to end. It all begins with your release strategy. Ideally, the way you release shouldn't rely on outdated run books, fragmented dashboards, and the instincts of that one engineer who's been here longer than anyone else. You know who I'm talking about. Even

Release at your speed with runtime control

as AI has significantly increase the speed of code generation, Launch Darkly enables releases to happen at your speed and allow you to maintain complete control at runtime. Need to make a big splash in an event? With one click, your software updates simultaneously across all of your mobile, browser, edge, and backend applications around the world in a fraction of a second. And if something

Instant rollbacks and safe release practices

goes wrong, which I'm sure will never happen to you, roll back just as fast. Day-to-day releases are different, though. They should be safe, observable, repeatable, and automated wherever possible. This becomes even more critical with the increase of AI generated code. That means exposing new functionality to small low-risk audiences first, measuring the effect of that change, and automatically and gradually increasing the targeted audience of that change. And here's the

Progressive rollouts and automated safeguards

best part. If anything degrades as the result of that release, LaunchDarkly can automatically detect the offending change, notify you, and even automatically pause it or roll back. There's no redeploy, no guessing. It's just safe by default. We're able to do this because LaunchDarkly provides comprehensive observability natively enriched with your feature release data in one place. That means your applications don't just tell you when a change regressed, they'll

Observability and self-healing systems

self-heal from failures. By the way, everything that we are talking about in this video applies to traditional software releases as well as those tricky AI agents that you've been trying to get into production. You can do all the evals you want, but unless you have a way to statistically analyze and respond to AI that interacts with your users, eventually it will cause problems.

Managing AI systems in production

Treating your AI configurations like static code is a mistake. AI is nondeterministic behavior that can't be unit tested against everything your chaotic users will throw at it. So, real time regression analysis in production paired with runtime kill switches are the best way to ensure that you're not shipping an anthropomorphic time bomb. When things drift, you don't just detect it, you contain it immediately. This is why Launch Darkly includes judges to continuously observe

Real-time regression detection and kill switches

and score your production AI implementation right alongside all your other telemetry. Sure, we've already contained the fire, but the next step is to figure out why that fire started in the first place. Here at Launch Darkly, we take the radical view that siloed data is bad because it leads to solving siloed problems. That's why we ingest all your user and application

Unified telemetry and end-to-end visibility

telemetry and cohesively stitch it together for a complete end-to-end view. This is the flow you already know and love. Set alerts, make dashboards, and follow traces across user sessions and distributed services. However, that's not enough for software delivery going at next-gen speeds. You need to get signal out of the noise automatically without needing to dig for it every

Release-level observability and signal detection

time. That's why Launch Darkly attaches relevant context to your telemetry data. On every flag, LaunchDarkly pre-filters observability data like errors, logs, traces, and sessions. And that means you get a full view not just of your system health, but of your release health at the feature level. You'll know instantly whether your release is working well or causing problems, so you can respond appropriately from that same flag that's already controlling the release.

Moving beyond firefighting to innovation

Obviously though, the goal is to move beyond firefighting so you actually get to build cool new stuff that makes the world better. So far, we've been focusing on derisking your changes so you get uninterrupted sleep. That's a great start. With good self-healing software in place, though, you're free to experiment and optimize your applications faster than ever. Let's be real,

Experimentation without bottlenecks

experimentation is hard. It takes careful setup, consistency of data, and all too often awkward handoffs across teams. Launch Darkly takes a better approach. Optimizing the end-user experience is everyone's goal, or at least it should be. And our entire platform is designed to bring your teams together around that effort. No more separation between engineering and business metrics. Use the same methods to get all the signals your teams need to know whether

Turning every feature into an experiment

your features are performant and driving the outcomes they are intended to. Because every flag or configuration can become an experiment, you start seeing results in minutes. And as soon as you're confident in a winner, you can instantly ship the right experience to your users. Don't leave your valuable features just sitting on the shelf. There's one last critical component that

Built-in AI for insights and decision support

we should talk about. Sure, you can measure everything and control anything, but knowing where to direct your efforts takes analysis and intelligence. That's why Launch Darkly includes built-in AI assistance where you can probably use a hand that helps you get insights into the raw data you're collecting to help you make the right decisions faster. So whether it's knowing how to

Closing the loop between insight and action

fix an underlying problem that LaunchDarkly already rolled back for you or helping think through how to best set up your next experiment, we're here to ensure that you can stay in the flow and continue to iterate. Moving fast without insight is risky and insight without the ability to do anything about it is just frustrating noise. That's why we think you should be able to measure what you control and control what you measure. That is the launch darkly way. You'll need rapid control and pertinent telemetry to deliver safely, on time, and with minimal stress. That feedback loop is the core way you can accelerate without introducing risk by building the right things for the right people. And LaunchDarkly gives you the fastest feedback loop there is,

Final thoughts and next steps

all at the feature level. Move fast, stay in control, and sleep at night. Sounds pretty good to me. Book a demo or start your free trial over at LaunchDarkley. com. Thank you.

Другие видео автора — LaunchDarkly

Ctrl+V

Экстракт Знаний в Telegram

Экстракты и дистилляты из лучших YouTube-каналов — сразу после публикации.

Подписаться

Лучшие методички за неделю — каждый понедельник