Figma Is Being Eaten
Not by another design tool — by designers who code. Nick Budden, a design-oriented entrepreneur who’s built multiple startups from Asia, sees dogma cracking at the fringes: “Designers are skipping Figma entirely. They’re vibe-coding directly to working prototypes.”
Tools are accelerating this. Cursor 2.2 now enables near-WYSIWYG website creation — designers manipulating live code visually, no handoff required. The Figma-to-engineer pipeline, a decade-old workflow, is compressing into a single seat.
But greenfield is the easy part. For mature products, the story’s more nuanced. Teams (like Notion) still run traditional design workflows — but they may not need to. The existing design already lives in code. That’s not a barrier; it’s a starting point. Designers with access to the codebase can iterate directly on what’s already built.
The real bottleneck isn’t tooling. It’s context.
The context bottleneck
Ask an AI to build a landing page from scratch — shockingly good. Ask it to modify your company’s checkout flow — endless loops of “that’s not what I meant”.
The difference isn’t model capability. It’s implicit knowledge: your pricing logic, that Q2 edge case, the integration your CTO rejected and why. None of it exists anywhere an agent can read.
Tools are chipping away at this. Kiro grounds generation in product specs. Glean surfaces documents. But both hit the same wall: company knowledge is scattered, contradictory, and stale. Search doesn’t fix structure.
What’s missing: Glean for agents
Nick’s hypothesis: AI-native firms need structured knowledge graphs that agents can reason over. Not document search. A system that understands relationships, detects inconsistencies, and surfaces conflicts for humans to resolve (like engineers merging code).
The immediate payoff isn’t even agents. It’s avoiding the meeting where people debate a project that was killed last week.
The diagnostic for leaders
Before scaling AI agents: Is your implicit knowledge encoded anywhere an agent could use it?
If not, you’re investing in capability without context. And you’ll keep hearing “that’s not what I meant”.