Isolated AI workflows create team disconnects. Pablo Stanley argues for integrating agents into shared, Slack-like environments where they become first-class participants. This allows for transparent, collaborative work between humans and AI, rather than having individuals work with agents in private.
Feeling his brain becoming addicted to AI's ease, Pablo Stanley deliberately learned pixel art by hand, resisting the constant urge to build an AI tool for it instead. This intentional, manual effort is a necessary exercise to protect one's creativity and avoid over-delegating thinking to models.
When pitching a complex new idea, Pablo Stanley goes beyond a simple screen share. He creates compelling async videos (Looms) where he tells a story, using hand-drawn maps and sketches to walk stakeholders through his thinking and build buy-in for his vision.
Pablo Stanley's design tool, Effecto, struggled using a 3D canvas that confused its LLM. It only became 'magic' when he rebuilt it using basic HTML and Tailwind, the native language of the models he was working with, dramatically simplifying development.
Stanley stopped developing his advanced design tool, Effecto, after realizing he wasn't using it himself. His personal workflow had evolved to simpler tools like TL Draw paired with AI agents, revealing a critical product-founder disconnect despite the tool being feature-rich.
The feeling of being a mere 'orchestrator' when using AI is not new. Pablo Stanley compares this creative crisis to artists like Andy Warhol, who ran 'The Factory' and directed others to produce his art. Accepting this shift from 'maker' to 'director' is a necessary mental model for modern creators.
The designer's role has shifted from Figma mockups to coding. At Vercel, designers are top code contributors, creating massive, exploratory pull requests on the production codebase. Engineers then act as 'refiners,' cherry-picking from these prototypes to ship polished, production-ready code.
With AI agents that understand design systems, designers no longer need pixel-perfect mockups. Instead, they provide 'ridiculously fast wireframes' with tools like TL Draw, then instruct the AI to apply high-fidelity styles, dramatically speeding up the design process.
The traditional master-apprentice model involved seniors teaching juniors by delegating tasks. Now, that work is delegated to LLMs. This starves junior designers of crucial learning opportunities, as seniors are now mentoring AI models instead of people, threatening the next generation of talent.