Long-form writing from the team — eval methodology, agent reliability, the economics of design-system labor, and what we ship next.
How we built 4,200 graded Figma → React pairs and what they reveal about today's models.
May 24, 2026 · 14 min read
EngineeringWhy we separate the executor from the verifier, and why that delta drives 94% merge rate.
May 12, 2026 · 9 min read
EssayIf LLMs write your code, your design system becomes the constraint that defines correctness.
Apr 28, 2026 · 7 min read
CustomerA play-by-play of a supervised brand migration that would normally take a quarter.
Apr 14, 2026 · 11 min read
EngineeringThe data model that lets us edit code and have Figma update in place.
Apr 02, 2026 · 12 min read
EssayAligning revenue with replaced labor cost — the only honest model for AI-native software.
Mar 22, 2026 · 6 min read