The Critical Advantage With AI
summary
Claims that the critical advantage is trainable skill — taste defined as tacitly judging quality against external standards — and that taste improves with immersion, valid cues, corrective feedback. Demonstrates an MCP Apps extension: server.registerResource linking a resource URI serving HTML with MIME type text/html; profile=MCP app, globals exposing tool_results, and a tunneled local MCP server rendered inside ChatGPT. Live demo shows an interactive Agent Conf schedule replacing “soulless” text and the UI working end to end.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 1:14The critical advantage in agentic engineering
- 2:27The answer is taste
- 3:39A scientific way to think about taste
- 4:08Defining taste
- 6:12Taste is trainable
- 6:22Ingredient one: immersion
- 8:02Ingredient two: valid cues
- 9:22Ingredient three
- 9:43Corrective feedback
- 10:58Recap and a Pixar example
- 11:59Practical coding demo
- 12:13A functional but poor web experience
- 14:02Reducing friction in agentic apps
- 14:50When plain text loses the brand
- 15:30MCP Apps
- 15:54Servers, tools, and resources
- 18:01Live demo in ChatGPT
- 19:53The Agent Conf MCP App
- 21:09Summary
related talks
33mFrontend after AI: The new UX
23mMCP-UI: Where Intent Becomes Interface
22mMCP UI: Extending the frontier
20mAI Agent Skills and MCP: Better Together
20mLetting AI Interface with your App with MCP
37mLetting AI Interface with Your App with MCPs
38mEvolving AI chat with MCP Apps
12mMCP and the Dawn of Agentic Intelligence