Dive Club 🤿 The trick to AI prototyping with your design system
31 snips
Jan 2, 2026 Lewis Healey, a design technologist at Atlassian specializing in AI prototyping, joins Kylor Hall, a design systems engineer, to explore scaling AI in established companies. They discuss the creation of pre-coded templates to enhance prototype consistency and strategies to reduce AI hallucinations. Hear about their innovative approach to simplifying design workflows and adapting roles in design systems. Plus, learn about benchmarking, engaging training tactics for non-technical users, and their vision for AI-native design systems that empower anyone to ship software.
AI Snips
Chapters
Transcript
Episode notes
Give Copyable Commands For Common Edits
- Provide copyable configuration commands for common changes (e.g., switch logo to Jira) to reduce hallucinations.
- Pair those commands with a constants object that maps product icons and elements to avoid incorrect substitutions.
Recipes Let Non-Engineers Add Features
- They created "recipes": small code blobs users paste to add features like a chat box or dark mode into existing prototypes.
- Recipes saved users from duplicating full templates and lowered technical friction for non-engineers.
Make The Model Think In Familiar Code
- Teach the LM to think in widely represented code patterns (like Tailwind) so it maps to your components reliably.
- Include a Tailwind-to-React translation to convert familiar classes into your internal component code and reduce errors.