As ChatGPT marks the third anniversary of its release, it's clear that the past three years have brought unprecedented disruption to the foundations of our digital ecosystems. We've seen massive consumer adoption—even if real productivity gains and large-scale production use still generate debate—but also a redirection of investment flows towards huge compute infrastructure, the challenging of long-standing monetization models by new gatekeepers, and a real questioning of the traditional SaaS playbook.
This period has also introduced a new semantic layer—copilots, agents, etc.—along with UX patterns that now seem here to stay (and conversation everywhere).
Among them, one has become increasingly visible over the past year while remaining surprisingly under-discussed: the canvas. I'm actually surprised that so few people are paying attention to this UX pattern, born from conversational interfaces, even though it signals a real shift in paradigm.
And, as often, Anthropic was the first to surface this intuition, launching "Artifacts" just one year ago, quickly followed by Google's AI Mode Canvas and ChatGPT's own Canvas. All three converge toward the same idea: an interactive workspace for writing, coding and prototyping—far beyond a simple chat interface. This signals a deeper architectural change, especially as app SDKs bring external capabilities directly into the conversational flow.
For decades, digital productivity had been structured around a fixed suite of applications—spreadsheets, text pads, slide editors and a few others. Team messaging platforms and visionary tools like Notion had already challenged that model.
But canvas-like interfaces take it a step further, as they introduce a reversal of perspective on who is the OS and who is the add-on: now more and more the AI stands as "the operating layer," while the app becomes an ephemeral sandbox anchored in a persistent context.
A quiet but consequential shift in how digital productivity is being redefined.
