Deepnote CEO: Why Notebooks Are the ‘Perfect User Interface’ for AI Agents

When Jakob Jurových and his team started building Deepnote in 2019, they saw a gap between two computational worlds that no existing tool could bridge. “World A was the world of tools that were simple to use, easy to get started with, but they also take you only so far,” said Jurových, Deepnote’s founder and CEO, pointing to spreadsheets as the prime example. “In World B, where tools are much more advanced, you can build anything that you can imagine — but first you need to spend a lot of time learning the tool.” Jurových made it his mission to build that missing computational middle ground. In this On the Road episode of The New Stack Makers, Jurových sat down with TNS Editor in Chief Heather Joslyn at JupyterCon in San Diego to talk about why Deepnote is going open source and to share his vision for notebooks as the ultimate medium for the AI era. The Missing Computational Medium Jurových liked the idea of notebooks, which had been around since the ‘80s, but the existing formats weren’t designed for the tight feedback loops that data exploration demands. Unlike software engineers with neat Jira tickets and clear end states, data scientists get handed a CSV file and told to “go find something interesting.” “Data exploration is a completely different way of working,” he said. “There’s no obvious endpoint, you can always go wider or deeper with data.” Deepnote was designed to be a notebook for constant collaboration, not asynchronous pull requests. “We showed how there can be not just two or three people pair programming in one notebook, but hundreds of people, all at the same time. And now we are routinely having sessions with thousands of people in one notebook.” From Scratchpad to Production The decision to go open source wasn’t simple. The team wanted to do it from day one, but had to solve more pressing problems first: stability, reproducibility and collaboration. “We realized that it’s important to solve the problems first, and then open source can be just the cherry on top,” Jurových noted. The team members also needed confidence in their architecture before committing to backward compatibility. Now, six years later, Deepnote is ready to go open source with a new format designed for the cloud, collaboration and AI era. Where Jupyter had two cell types (code and markdown), Deepnote has 23 building blocks — and counting. “We see notebooks as a beautiful format where you can actually stay and keep in the same place all the way to productionizing your workflow,” Jurových said. “The notebook itself can become the whole data app. It can become that thing that you schedule to run every 12 hours. It can have an API endpoint attached to it.” This flexible multi-tasking capacity is why, he concluded, “Notebooks are the perfect user interface for working alongside AI agents.” Check out the full episode to hear more about Deepnote’s journey from breaking under classroom load to a production-ready tool, and why the next decade of computational work might look nothing like the formats we’ve relied on since 2011. The post Deepnote CEO: Why Notebooks Are the ‘Perfect User Interface’ for AI Agents appeared first on The New Stack.