Just shipped Canvas and Video editors for AOSE — an open-source office suite where AI agents are first-class collaborators.
Canvas is a Figma-like infinite canvas editor with vector editing, pen tool, boolean operations, and nested grouping. Export to PNG/SVG.
Video is a timeline-based motion graphics editor with keyframe animation. Export to MP4/WebM.
Both are built on HTML — which means agents aren't filling templates. They have the full expressive power of the web: SVG, CSS, keyframe animation, all directly accessible.
Humans don't need to touch HTML. A projection engine maps editable properties from the HTML — position, color, size, animation timing — onto the UI. Anything the AI writes into HTML that isn't mapped out can only be changed by the AI. The range of what humans can control grows as the projection layer expands.
AOSE now has 6 editors: Docs, Databases, Slides, Flowcharts, Canvas, and Video. Humans and agents edit the same object.
