When we started building Zerve, we had a couple of perspectives on the space that we learned both through our own experience and through hundreds of interviews we performed over about a year. First, the tools available to data scientists force them to choose between stability and flexibility. Exploring data in VS Code is a struggle, and deploying a notebook is nearly impossible. Second, the only people generating value from data today are doing it with code. In other words, 100% of AI projects require code in some way.

That’s why we built Zerve, and that’s why we made it free for our users.

We launched on January 30, and in the first week had thousands of people sign up to use Zerve. The feedback was immediate and valuable, confirming that we’re on the right track. Here are some of the lessons we’ve learned since the launch.

The Community Recognizes the Problem

Every user who signed up for Zerve received a direct email from me. Our goal was to hear about their experiences firsthand. We had already run a beta test in 2023, but the launch validated what we suspected: people are frustrated with current tools.

We heard feedback like, “Zerve AI is like Miro for developers, except cooler,” and “It’s like the Google Colab upgrade we’ve all been waiting for.” These reactions matched our vision perfectly, showing us that the problem and solution were clear to users.

Since the viral 2018 talk “I Don’t Like Notebooks”, the data science community has been aware of notebook limitations. Our launch added momentum to that recognition.

Self-Hosting Is Critical for Data Science

Most tools in the last decade were built in the cloud, but users consistently tell us they want to keep data inside their own environments. No one wants sensitive data leaving their cloud or being processed by a third party. Vendors shouldn’t want that either—it introduces security and compliance risks.

That’s why Zerve was designed to be self-hosted from the beginning. We’re on the AWS Marketplace, and setup takes under 10 minutes. While we offer a cloud option, conversations with users confirm that self-hosted deployments are essential for real projects on real data.

We’ve also heard requests for local installs, but our focus is on streamlining the path from prototype to production with collaboration and cloud-first technologies. This approach makes handoffs and deployment far more efficient.

Feedback on Bugs and Gaps

Like any launch, there have been bugs. What stood out was the patience and positivity of our community. Users flagged gaps, both expected and unexpected, and those conversations helped us refine the product. It’s been invaluable to meet users, understand their motivations, and align functionality with our roadmap.

Generative AI Is Everywhere

Initially, I thought generative AI might be hype. I was wrong. Nearly every organization we speak with is working on LLMs or other generative AI projects. These models represent a leap forward, but the tooling for training and deploying them is complex and costly.

Zerve helps by providing serverless GPU support, removing common bottlenecks with compute and infrastructure management. Next month, we’ll roll out new features, including integrations with Hugging Face and AWS Bedrock, plus expanded GPU support. These updates will strengthen Zerve’s position as the IDE for generative AI.

Stay tuned for updates on our website and Product Hunt page.