Deepnote Ceo: Why Notebooks Are The ‘perfect User Interface’ For Ai Agents

When Jakob Jurových and his squad started building Deepnote successful 2019, they saw a spread betwixt 2 computational worlds that nary existing instrumentality could bridge.

“World A was nan world of devices that were elemental to use, easy to get started with, but they besides return you only truthful far,” said Jurových, Deepnote’s laminitis and CEO, pointing to spreadsheets arsenic nan premier example.

“In World B, wherever devices are overmuch much advanced, you tin build thing that you tin ideate — but first you request to walk a batch of clip learning nan tool.”

Jurových made it his ngo to build that missing computational mediate ground.

In this On nan Road section of The New Stack Makers, Jurových sat down pinch TNS Editor successful Chief Heather Joslyn astatine JupyterCon successful San Diego to talk astir why Deepnote is going unfastened source and to stock his imagination for notebooks arsenic nan eventual mean for nan AI era.

The Missing Computational Medium

Jurových liked nan thought of notebooks, which had been astir since nan ‘80s, but nan existing formats weren’t designed for nan tight feedback loops that information exploration demands. Unlike package engineers pinch neat Jira tickets and clear extremity states, information scientists get handed a CSV record and told to “go find thing interesting.”

“Data exploration is simply a wholly different measurement of working,” he said. “There’s nary evident endpoint, you tin ever spell wider aliases deeper pinch data.”

Deepnote was designed to beryllium a notebook for changeless collaboration, not asynchronous propulsion requests. “We showed really location tin beryllium not conscionable 2 aliases 3 group brace programming successful 1 notebook, but hundreds of people, each astatine nan aforesaid time. And now we are routinely having sessions pinch thousands of group successful 1 notebook.”

From Scratchpad to Production

The determination to spell unfastened root wasn’t simple. The squad wanted to do it from time one, but had to lick much pressing problems first: stability, reproducibility and collaboration.

“We realized that it’s important to lick nan problems first, and past unfastened root tin beryllium conscionable nan cherry connected top,” Jurových noted. The squad members besides needed assurance successful their architecture earlier committing to backward compatibility.

Now, six years later, Deepnote is fresh to spell unfastened source pinch a caller format designed for nan cloud, collaboration and AI era. Where Jupyter had 2 compartment types (code and markdown), Deepnote has 23 building blocks — and counting.

“We spot notebooks arsenic a beautiful format wherever you tin really enactment and support successful nan aforesaid spot each nan measurement to productionizing your workflow,” Jurových said. “The notebook itself tin go nan full information app. It tin go that point that you schedule to tally each 12 hours. It tin person an API endpoint attached to it.”

This elastic multi-tasking capacity is why, he concluded, “Notebooks are nan cleanable personification interface for moving alongside AI agents.”

Check retired nan afloat section to perceive much astir Deepnote’s travel from breaking nether schoolroom load to a production-ready tool, and why nan adjacent decade of computational activity mightiness look thing for illustration nan formats we’ve relied connected since 2011.

Group Created pinch Sketch.