Quodra

Technical work is interactive, computational and dynamic, but the tools used to present it are still largely static. The moment an equation, a circuit or a chart goes into a slide it becomes a picture: you cannot change a variable, re-plot the data, or ask an assistant to adjust it without going back to whatever produced it. Quodra is built on the opposite assumption. Objects are the truth; slides are a projection.

←

→1 / 3

1. 01The editor
2. 02A slider that drives six objects
3. 03Every connection in the deck

Objects, not pictures

Every object in Quodra has a typed representation that the editor, the backend and the AI interface share:

• Equations are LaTeX trees you keep editing, not renders, and they can compute.
• Python cells run in the browser through Pyodide, with output attached to the slide.
• Charts and data tables are bound together, so changing the data re-plots the chart.
• Circuits are real schematics. Circuit Studio solves DC, AC and transient analyses, so a circuit on a slide is solved, not drawn.
• Molecules come from SMILES strings, and Matter Lab adds crystal structures, phase diagrams, diffraction patterns, stress–strain curves and Ashby charts.
• Simulations, diagrams and citations follow the same pattern.

←

→1 / 11

1. 01An equation, still a LaTeX tree
2. 02A Python cell, run in the browser
3. 03A chart bound to its table
4. 04Circuit Studio: the operating point, solved
5. 05The same circuit, swept in frequency
6. 06Molecules from SMILES
7. 07The Fe–C diagram, with a live lever rule
8. 08Two crystal structures of iron
9. 09Diffraction, simulated from the cell
10. 10An Ashby chart with a design line
11. 11A trajectory, played on the slide

I designed the document model before the editor. That order is what keeps AI-generated content editable instead of pasted in as an image, and it is what let realtime co-editing, a presenter view and audience windows arrive later without changing what a slide is.

An interface for assistants

Quodra speaks MCP, the Model Context Protocol, so an AI assistant can read decks, create new ones and edit them using the same vocabulary the editor uses. It is not pasting text into a template: it places objects, sets parameter ranges and wires a slider to the thing it drives, and a batch of edits lands whole or not at all. The same object model that keeps human edits structured keeps machine edits reviewable and reversible. Quodra deliberately provides no built-in model; the client brings its own.

Meeting people where their files are

Decks still have to leave the app. Quodra imports PowerPoint and Google Slides, and exports PDF, PowerPoint, images, its own JSON and an embeddable viewer, so existing material can come in and finished work can go out to whatever a lecture hall or a company already runs on.

How it is built and run

The editor is React 19 and TypeScript on Vite. The API is NestJS on Google Cloud Run in front of Cloud SQL Postgres with row-level security, and the static surfaces and the hosted MCP endpoint run on Cloudflare. The codebase carries 423 unit tests plus Playwright end-to-end tests, and every change goes through CI before staging. Quodra grew out of Praxis, an earlier object-centric prototype I built at the end of 2025, and is now in private beta with a public sample anyone can explore. I lead product and engineering alongside a co-founder who leads design and a team of five in total.