GrowthBook
The product development platform for AI-native teams.
Start new thread

GrowthBook 5.0 - Build, ship, and improve at scale

by•
GrowthBook 5.0 brings feature flags, experimentation, and product analytics into one AI-native, warehouse-native platform. Build no-code experiments in the browser with the new AI Visual Editor, let agents create flags and draft experiments with 25 open-source Skills, explore product data through the in-app AI Assistant, and ship safely with stronger governance. Faster queries and a streamlined experiment workflow help every team move from idea to insight with less friction.

Add a comment

Replies

Best
Maker
📌
Hey Product Hunt! We’ve spent the past year rebuilding and expanding nearly every part of GrowthBook, and 5.0 brings all of that work together. AI has made it much easier to build and ship software. We wanted to make it just as easy to test ideas, learn what works, and roll changes out safely without losing the rigor and control experimentation teams depend on. GrowthBook 5.0 includes a new AI-first Visual Editor for building no-code experiments directly in the browser, 25 open-source Skills that let agents work with feature flags and experiments, an in-app AI Assistant, and Product Analytics, now generally available. We also added stronger feature flag governance, faster queries, and a much simpler experiment workflow with about 20 fewer form fields. The goal is to give humans and agents one consistent place to move from an idea to an experiment to an insight, with fewer handoffs and less busywork. We’d love to hear what you think, what you want to try first, and what you’d like us to build next. Check out our Office Hours Video to see our Founders and Engineers who worked on the 5.0 release go into their favorite features:

 We hit the classic problem where the experimentation tool was solid but nobody outside the growth team could read the results, so calls still happened on vibes. What did 5.0 change on the interpretation side for non-technical stakeholders? Flags are easy, getting a PM to trust the readout is the real work.

Maker

 

Yeah, this is exactly the hard part. Running the experiment is one thing. Getting someone outside the growth or data team to look at the result, understand it, and actually trust what they are seeing is the much bigger job.

I would not say 5.0 magically solved that, but it did make a lot more of the interpretation work self-serve.

You can use the in-app chat to ask questions about an experiment without having to know how to read every part of the results view first. You can ask what changed, what looks meaningful, or where you should dig deeper. Experiment dashboards also make it easier to bring the result, the supporting metrics, and the surrounding context into one place instead of sending someone through five different screens and hoping they piece it together.

And because 5.0 is much more agent-friendly, you can also use the AI tool you already work in to help analyze results, explain the readout, or ask follow-up questions in more natural language. So it is less “here is the stats page, good luck” and more “here are a few different ways to understand what happened.”

That does not remove the need for good metric definitions or healthy skepticism. But it does mean a PM or stakeholder can get much further on their own before they need an analyst to translate everything for them. That is really the shift in 5.0: making experimentation easier to understand, easier to explore, and a lot more self-serve.

There’s a much deeper walkthrough here if you want to dig in: