What if your AI assistant could watch you code, spot bugs as they happen and fix them before you even ask? That’s the future David Qiu, Jupyter Security Council member and software engineer at Amazon Web Services, envisions for Jupyter AI — and it’s closer than you might think.In this On the Road episode of The New Stack Makers, Qiu and Piyush Jain, principal engineer at AWS and Jupyter distinguished contributor, sat down with TNS Editor in Chief Heather Joslyn at JupyterCon in San Diego to discuss the just-released Jupyter AI v3 and what’s ahead on the roadmap.What Are AI Personas?A key addition to v3 is AI personas: customizable AI assistant configurations within the Jupyter AI extension for JupyterLab. These personas are specialized agents that can use different models and frameworks to help users with various tasks directly within the Jupyter environment.“Unlike many other AI extensions out there, Jupyter AI uniquely allows for multiple named agents — AI personas — to be available inside of every chat,” said Qiu, who has been a Jupyter AI maintainer since its debut at Jupytercon 2023. “Users get to actually select which agents they want to invoke, like a Claude Code persona or an OpenAI Codex persona.”Developers can also build their own personas as local AI agents that load live into JupyterLab chat, or as a distributable package that other developers can install via pip like any other Python package.What Jupyter AI’s Codebase Split Made PossibleAI personas only became possible when Jupyter AI’s massive original codebase was split into multiple smaller, more manageable packages for v3.“Contributors were having a really hard time understanding the code, which piece did what job,” said Jain. “Now, you can pick which packages you want to install on your environment.”Users can also swap out core components entirely, Qiu noted: “Consider a use case where an enterprise wants to have a custom message router. In v3, they can just uninstall our router package and provide their own.”Looking ahead, Qiu envisions Jupyter AI evolving into “an ecosystem of AI personas” that enables true multiagent collaboration.“Imagine having access to a whole team of agents,” he said, describing a future where a high-level persona could dispatch requests to specialist agents. “You could have a data science persona, software engineer personas, plus QA and tester personas, and actually orchestrate it just like a software manager.”With representatives from AWS, Quansight, Apple, and connections to stakeholders at Meta and Netflix, the Jupyter AI community is poised for growth. The project’s modular architecture makes it easier than ever for organizations to contribute custom personas and specialized packages back to the broader ecosystem.Check out the full episode to learn more about Jupyter server documents, how Jupyter AI v3’s performance optimizations work under the hood and the roadmap for what’s next.The post Jupyter AI v3: Could It Generate an ‘Ecosystem of AI Personas?’ appeared first on The New Stack.