Dive Club 🤿 Brian Lovin - How to level up with AI as a designer
159 snips
Apr 14, 2026 
Brian Lovin, product and design leader who built AI prototyping at Notion and created the read-it-later app Shiori. He discusses shifting design workflows into code and prototype playgrounds. He covers using multiple models, tooling like Agentation and Conductor, prompt craft, hardening AI systems, and strategies for designers to stay relevant as AI changes how work gets shipped.
AI Snips
Chapters
Transcript
Episode notes
Last Mile Polish Still Needs Humans Or Rigorous Primitives
- AI is weak at last-mile visual polish, so teams either tweak CSS by hand or invest in high-fidelity Figma primitives for AI to reuse.
- Naming layers/components consistently makes Figma more legible to models and improves automated conversions to code.
Make Intent Legible With Component Names
- Make your intent legible to AI by referencing component names, file locations, or IDs instead of vague visual descriptions.
- Tools like Agentation or DialKit increase fidelity because the AI can act on explicit component identifiers.
Experiment Widely To Learn Each Tool's Strengths
- Try many AI tools non-tribally and learn which model excels at which task by experimenting.
- Brian switches between Conductor, Cursor, Cloud Code and different models (Opus, Codex, Claude) based on speed and task type.