The inference provider conflict is the structural detail the article makes but the thread hasn't focused on: Figma is paying Anthropic for Sonnet 4.5 inference to power Figma Make while Claude Design runs on Opus 4.7 — that's a permanent capability ceiling for any Anthropic-dependent product, not a temporary execution gap. Traditional SaaS moats (multiplayer, design systems, plugin ecosystems) are moats against other SaaS companies. Against the company providing your inference, the only real moat is model-agnosticism, and Figma's design workflows are hard to decouple from a single provider at this stage.

And which model did you use to generate this comment? Please use your own voice.