What's New

Source Traceability

Published briefs now show where every visualization comes from. Each chart, table, and image includes a "View source" link that reveals the code behind it—so readers never have to wonder how a number was calculated.

One-click provenance: Click "View source" on any artifact in a published brief to see the exact code that produced it. Click through to open the full notebook, which scrolls to and highlights the source cell.

Works across notebooks: Briefs that pull artifacts from multiple notebooks track each source independently. Every link goes to the right notebook and the right cell, regardless of where it came from.

Built for trust: When a stakeholder questions a chart, the answer is one click away. No searching through notebooks, no asking the author—the source is right there.

Smarter Sharing

Sharing now gives you more control and gives viewers more context.

See what's connected: Public briefs and notebooks now show which datasets were used in the analysis. Viewers can click through to explore the data, understanding where your insights come from.

One share modal, everywhere: Sharing a brief, notebook, or dataset now works the same way everywhere. Toggle visibility, copy the link, and see what's connected—all in one place.

Dataset tracking: Mark which datasets your notebook uses. When you share the notebook (or a brief linked to it), viewers see those datasets listed with links to explore them.

Quick access to public view: A new "View public page" button in the editor lets you preview exactly what others will see when they visit your shared link.

Faster Onboarding

New notebooks now start with a welcome cell that gets you working immediately.

Ready-to-run starter code: Every new notebook includes imports for pandas, numpy, and matplotlib—plus a table showing your available datasets and how to load them. Press Shift+Enter and you're off.

Learn the shortcuts: The welcome cell introduces the AI assistant shortcut (⌘J / Ctrl+J), so you know help is one keystroke away.

Customize or disable: Don't want the starter code? You can customize the welcome cell content in settings or turn it off entirely. Your notebooks, your workflow.

Public Sharing

Your work can now be shared publicly. Generate a shareable link for briefs, notebooks, or datasets and anyone can view them—no login required.

What you can share:

• Briefs – Publish polished reports with interactive visualizations
• Notebooks – Share your analysis with full code and outputs visible
• Datasets – Let others preview and download your data

How it works: Open the share modal on any brief, notebook, or dataset and toggle public access. You'll get a link you can send to colleagues, embed in documentation, or post anywhere. Public items appear on your profile at /@yourname, building a portfolio of your work.

AI Assistant

There's now an AI assistant built into the notebook. Open it from the sidebar, describe what you're trying to do, and it writes and runs Python code on your behalf.

How it works: The assistant sees your notebook context—existing cells, variables, and outputs—and can add new cells, modify code, or execute analysis. It uses the same kernel as you, so results appear directly in your notebook.

Conversational sessions: Each conversation is a session you can return to. The assistant remembers what you've discussed and builds on previous work. Start fresh or continue where you left off.

Notebook UX Improvements

Major improvements to the notebook experience, making it faster and more intuitive to start working.

Auto-connect kernel: Kernels now connect automatically when you open a notebook. No more clicking "Connect kernel" before running code—just open and start working.

Execution queue: Run cells anytime, even while another cell is executing or before the kernel is ready. Cells queue up and execute in order, just like Jupyter. Visual indicators show which cells are queued, running, or complete.

Instant notebook creation: Clicking "Blank notebook" now creates and opens the notebook immediately. No intermediate save step—auto-save works from your first keystroke.

Full import/export:

• Import preserves all outputs including charts, images, tables, and errors
• Export downloads a standard .ipynb file compatible with JupyterLab, Colab, and Kaggle
• Round-trip fidelity: export from Margin and open anywhere

First-time guidance: New users see a helpful hint explaining that the kernel connects automatically and cells can be queued. Dismiss it once and it won't appear again.

Better feedback:

• Cells queued while the kernel is connecting show a one-time toast explaining what's happening
• Clearer error messages for import failures
• Retry affordance if notebook creation fails

Profiles & Sharing

Claim your username and get a public profile at /@yourname. Publish briefs and notebooks with one click—shareable URLs with Open Graph previews included.

Stripe billing is live with a generous free tier. Monitor compute hours and storage from your dashboard.

Live Python Kernel

Your code now runs on real cloud infrastructure. Python 3.11 with pandas, numpy, and matplotlib pre-installed. Results stream back in real-time.

The editor uses CodeMirror 6 with kernel-backed autocomplete, Cmd+I inspection, multiple themes, and vim-like keyboard navigation. Markdown cells render LaTeX via KaTeX.

Cloud Infrastructure

Margin now has a real backend. Supabase powers authentication, notebooks, briefs, and datasets with real-time sync. Multi-tenant organization architecture and row-level security ensure data stays private.

Your work saves automatically. No more lost notebooks.

Development Begins

The first lines of code for Margin were written on Christmas Day. We set out to build a clean, approachable interface for data analysis—Jupyter-compatible notebooks that run in the cloud with first-class support for publishing and sharing.

Initial stack: Nuxt 4 with Nuxt UI, dashboard shell, and early landing page with product vision.