ReadingResearchRadarInvestment framework
Sign in / Sign up
← Back to reading
NewsMaterial3 min readDisconfirmAI Editorial

Itay Schiff, Co-founder of Figma Weave: In the Creative Process, It’s Labor That Should Be Handed to AI, Not Thought

Starting from his experience of turning to ChatGPT within ten minutes in a writing class, Itay Schiff proposes a specific criterion: before automating, ask whether a step is labor or thought. Using examples from VFX workflows and Gaussian splatting projects, he illustrates that generation does not equal authorship, and distinguishes between two modes of work: open mode and closed mode.

2026-11-14 04:00 GMT+8

Itay Schiff serves as co-founder and chief creative officer at Figma Weave, having worked for twenty-five years in post-production and visual effects. He recently enrolled in a writing class hoping to clarify his thoughts on the relationship between AI and creative work through writing. The first assignment was simple: choose a topic and write about it. He began jotting down notes, but within ten minutes found himself asking ChatGPT to complete the rest for him. Later reflecting on the experience, he realized his goal wasn’t to produce a finished piece, but to sort out his own ideas—yet he still couldn’t resist outsourcing his thinking.

This experience drew his attention to a broader tension: many creative professionals are using AI to do things they originally enjoyed doing—writing, designing, editing, directing. While speed has indeed increased, this comes with a deep sense of conflict. Schiff believes that debates about AI and creativity are often oversimplified into opposing camps of supporters versus detractors, but this division is too coarse. He is a daily user of AI and is also building AI products, yet he is increasingly worried about how easily he surrenders parts of the work that used to truly matter to him.

The core criterion he proposes is: before handing over any step in your process to AI, first ask what that step does for you. If it is merely labor—for example, shooting on a green screen, removing tracking points, building previs, adding 3D renders, pulling the green screen, and refining edges in a VFX workflow—these steps may take days or even weeks, but they do not actually contribute to the artistic outcome. In such cases, let the machine take over. However, if a step involves thinking—for instance, when manually editing a shot you are simultaneously considering its artistic impact, or when drafting a creative brief you are clarifying your vision—then it represents valuable friction worth protecting.

Sign in / Sign up
ReadingResearchRadarInvestment framework

Schiff further points out that AI generation does not equal authorship. The first version produced by a model often looks high-quality, unlike the rough drafts of the past, making it tempting to adopt directly. But anyone could generate the same image. He cites a promotional video Figma Weave recently created for its Gaussian splatting feature: the video, showcasing a 3D scene rendering tool, was generated by the Seedance 2.5 model. Technically, it could have been captured in a single continuous shot, but the thinking behind it took much longer. Starting from a few images and a prompt, the first version came out quickly but fell far short of their accepted standards. They repeatedly adjusted the story, characters, and style until the model’s output made them feel proud.

From the experience of building the Weave product, Schiff distilled two modes of creation: open mode, which involves exploration, producing first drafts, changing direction, and embracing uncertainty; and closed mode, which is structured, locks in decisions, automates tasks, and makes work repeatable. He believes creators and tools alike need to switch between these two modes, and that the most powerful state of a workflow is one that preserves choices a creator has already made while leaving room to reopen those choices when necessary.

He ultimately acknowledges that this boundary cannot be universally drawn: a step that feels like drudgery to one person might be creatively thoughtful to another; the same person may also change their judgment across different projects. Yet his practical advice remains: examine anything before handing it off to AI, and be especially wary of automating thinking itself in places where thinking is most easily replaced—namely, where decision-making is hardest and where the lack of ready-made answers causes the most discomfort.

Sources

  • figma.comFigma↗
Continue researching →