The Critical Advantage With AI

Kumar argues that having powerful tools does not guarantee high-quality work, and distinguishes craft from slop. He defines taste as a trainable human skill for quickly judging quality against external standards.

Kumar identifies three ingredients: immersion in high-quality examples, valid cues that connect first impressions to how something performs, and corrective feedback from someone more experienced. His computer mouse example shows why looking at many designs is not enough without using them and learning from informed critique.

Kumar contrasts the friction of ordinary websites with the ease of asking an agent for information, while noting that a plain text answer can lose a company’s branding and design. He demonstrates an MCP App that connects a schedule tool to an HTML resource, letting an MCP client render interactive, branded speaker and schedule details inside ChatGPT.

Kumar says that broad access to creation tools makes the quality of what people build a key differentiator. He encourages teams to apply immersion, valid cues, and corrective feedback, and mentions IBM’s coding agent, Bob, as an effort to make an agent more considered.