When execution gets easy, taste gets harder (www.antonsten.com)

AI-driven tooling is blurring the long-standing handoff between designers and developers: designers can now generate working CSS and interfaces by prompting tools like Cursor, while engineers routinely tweak UI details directly. The author argues this isn’t a takeover but an evolution—similar to how Figma shifted design toward systems thinking—but the next inflection will be tools that truly blend structured code with an editable visual canvas (Framer is flagged as a likely contender). That shift makes execution trivially available to more people; the technical novelty is less about raw generation and more about seamless bi-directional editing between visual and code representations. For the AI/ML community this matters because models and platforms will need to support mixed-mode workflows, preserve design intent, and enable higher-level judgement rather than just pixel output. As execution lowers in friction, the scarce competency becomes “taste”: judgmental pattern recognition that chooses which problems to solve, which solutions will work in context, and how designs align with user and business goals. Practically, designers should invest less in execution speed and more in strategy, research, and conviction; tool builders should focus on integrations that support human judgment, context-aware suggestions, and collaborative, editable code+visual representations rather than pure automation.