The New Stack Podcast Jupyter AI v3: Could It Generate an ‘Ecosystem of AI Personas’?
10 snips
Nov 12, 2025 David Qiu, a software engineer at AWS involved in Jupyter AI v3, and Piyush Jain, a principal software engineer at AWS, dive into the exciting developments in JupyterLab. They introduce customizable AI personas that help with tasks like coding and debugging, allowing users to create and share their own agents. The duo explains the benefits of a modularized codebase for flexible integration and collaboration features, envisioning an expansive ecosystem of AI personas that will enhance multi-user interactions and proactive AI support in workflows.
AI Snips
Chapters
Transcript
Episode notes
Multiple Named AI Personas In Chat
- Jupyter AI v3 introduces named AI personas that users can mention in chat to invoke specialized assistants.
- Multiple personas like "Claude Code" or "OpenAI Codex" can coexist and be selected within a single JupyterLab chat.
Build Local Then Package Personas
- Build personas locally to iterate quickly using the refresh personas slash command inside JupyterLab.
- Package personas as pip-installable distributions to share and replicate them across users and environments.
Modular Packages Replace Monolith
- Jupyter AI v3 split its monolithic codebase into modular packages to ease contribution and customization.
- Admins can now install or replace specific components like personas or message routers instead of editing a single large repo.