TL;DR
Figma's Make AI generator improves rapidly but struggles with code quality, determinism, and economics—success depends on becoming production-ready, not just prototype-fast.
Key Points
- Make defaults to Claude Sonnet 4.5 with experimental multi-model support (Gemini 3 Pro); outputs hosted on AWS/Cloudflare
- GitHub integration limited to one-way pushes to Make-created repos only—no two-way sync, branch management, or Enterprise Server support
- Community feedback: praised for speed and stakeholder alignment; criticized for generic aesthetics, brittle code, and opaque credit-based pricing moving to enforced limits in March 2026
- Competitive positioning: Make compresses design→prototype phase but doesn't solve backend systems, architecture, or production handoff—positioning it as demo-first rather than production-first
Why It Matters
Make represents Figma's bet to own the 'first runnable draft' in the software creation loop, but its constrained GitHub integration and generated code quality suggest it will remain a prototyping layer rather than a production tool. If Make can't reliably become the starting point for engineering teams, users will default to code-native AI workflows (Claude, Cursor, Anthropic) that operate deeper in the pipeline.
Source: chadsnewsletter.substack.com