Nearly Headless Web
Overview
The nearly headless web is a product architecture where most discovery, understanding, and action can be handled by agents through machine-operable surfaces, while selected moments still keep a human-visible interface for judgment, preference, review, or handoff.
Significance
Fully headless APIs are not always enough. Some user choices remain visual, comparative, or preference-heavy. Liad Yosef's agentic-web framing argues that the web is moving toward headless operation without becoming completely headless.
Applied Use
- Expose structured actions for routine work.
- Preserve human UI for ambiguous, high-risk, or taste-based decisions.
- Make handoff explicit: agent prepares, human reviews, agent executes or records.
- Pair MCP Apps, WebMCP, browser automation, and normal web UI based on the job.
Connections
Evidence Graph
This evidence graph consolidates scheduled talks, linked videos, transcripts, and slide-derived material connected to this topic.
Linked Sessions
Media Signals
youtube-o-zkvb0iFDQ— 3,969 transcript words; 1 slide-derived text signals- Transcript signals for
youtube-o-zkvb0iFDQ: apps, host, claude, back, chatgpt, look, mcpui, chat. - Slide-derived themes for
youtube-o-zkvb0iFDQ: engineering, future. - Evidence links for
youtube-o-zkvb0iFDQ: youtube o zkvb0iFDQ, youtube o zkvb0iFDQ transcript, youtube o zkvb0iFDQ slides, youtube o zkvb0iFDQ dense slides, youtube o zkvb0iFDQ reconstructed slides
Evidence Boundary
This page synthesizes a recurring design frame from the official schedule description and related transcript evidence. It should be refined if an exact session recording becomes available.