Markdown source

Recursive Model Improvement

Conference Context

• Date/time: 2026-06-29 · 5:10pm-5:30pm
• Track/room: Software Factories · Main Stage
• Speaker(s): Lee Robinson
• Session type/status: keynote · confirmed

• Track: Software Factories
• Room: Main Stage
• Session type: keynote
• Status: confirmed

Session Description

No official description published in the schedule data.

Media Evidence

Building Cursor Composer – Lee Robinson, Cursor (speaker-match related prior/adjacent AI Engineer video; captions: English auto-captions).

• youtube htM02KMNZnk transcript — full cached transcript markdown for the related YouTube source.

• Source video: youtube-fL1iJHtl51Q
• Slide deck: Dense Slides: Building Cursor Composer – Lee Robinson, Cursor — 2 visible slide image(s); 2 HTML recreation(s).



• Additional slide evidence: Slides: Building Cursor Composer – Lee Robinson, Cursor, Reconstructed Slides: Building Cursor Composer – Lee Robinson, Cursor
• Slide-derived themes for youtube-fL1iJHtl51Q: speed, composer, intelligence, frontier, combines, coding, score, tokens.
• Source video: youtube-htM02KMNZnk
• Slide deck: Dense Slides: WF2026: Software Factories & Keynotes ft. Microsoft, OpenAI, OpenClaw, Z.ai (GLM), MiniMax, HF — 18 visible slide image(s); 18 HTML recreation(s).



• Additional slide evidence: Slides: WF2026: Software Factories & Keynotes ft. Microsoft, OpenAI, OpenClaw, Z.ai (GLM), MiniMax, HF, Reconstructed Slides: WF2026: Software Factories & Keynotes ft. Microsoft, OpenAI, OpenClaw, Z.ai (GLM), MiniMax, HF
• Slide-derived themes for youtube-htM02KMNZnk: cycles, stacking, loops, tokens, tools, tasks, throughput, many.

• youtube q4Tr DknG2M transcript — full cached transcript markdown for the related YouTube source.

Evidence Graph

This evidence graph is generated from currently linked source material: official schedule text, related video pages, cached transcripts, visible slide text, dense/reconstructed slide pages, and AI slide-classification audits.

Media Signals

• youtube-fL1iJHtl51Q — 5 slide-derived text signals
• Slide-derived themes for youtube-fL1iJHtl51Q: speed, composer, intelligence, frontier, combines, coding, score, tokens.
• Evidence links for youtube-fL1iJHtl51Q: youtube fL1iJHtl51Q, youtube fL1iJHtl51Q slides, youtube fL1iJHtl51Q dense slides, youtube fL1iJHtl51Q reconstructed slides
• youtube-htM02KMNZnk — 89,050 transcript words; 4 slide-derived text signals
• Transcript signals for youtube-htM02KMNZnk: model, code, models, loop, well, software, first, team.
• Slide-derived themes for youtube-htM02KMNZnk: cycles, stacking, loops, tokens, tools, tasks, throughput, many.
• Evidence links for youtube-htM02KMNZnk: youtube htM02KMNZnk, youtube htM02KMNZnk transcript, youtube htM02KMNZnk slides, youtube htM02KMNZnk dense slides, youtube htM02KMNZnk reconstructed slides

Agent Reading Notes

Use these signals to refine the synopsis, topic links, people/company context, and method notes. If a source is a related external video rather than an exact official recording, keep it framed as supporting evidence.

Transcript Status

Related video transcript availability: English auto-captions. Treat this as supporting context, not a recording of this exact scheduled session unless later confirmed. Not fetched yet.

People

• lee robinson

Supporting Slides

• youtube fL1iJHtl51Q slides — extracted from the related public AI Engineer video.

Slide Evidence

• Slide-only cropped deck: youtube fL1iJHtl51Q dense slides (2 viable slide images).
• Related slide/OCR pages:
• youtube fL1iJHtl51Q dense slides
• youtube fL1iJHtl51Q reconstructed slides
• youtube fL1iJHtl51Q slides
• Slide-derived terms: frontier, file, composer, fast, cursor, intelligence, speed, best, model, robinson, developer, education, read, edit, code, combines, coding, best-in-class

Livestream Segment

• Watch in livestream at 08:12:13 — WF2026: Software Factories & Keynotes (Day 1).
• Match basis: speaker and title; timed captions matched Lee Robinson, improvement, model, recursive.
• Confidence: high automated match; prefer a dedicated cut-video recording when one exists.

Synthesis

Synthesized Breakdown

The session explains MCP Apps as a way to add interactive UI to MCP-based workflows. It treats the MCP host as the place where tool data, model reasoning, and human controls meet.

Speaker And Company Context

• Lee Robinson — ML, Model Behavior at Cursor.

Topics Covered

• agent security
• agentic search
• agentic web
• ai sandboxes
• coding agents
• mcp
• mcp apps

Derived Links And Source Material

• youtube htM02KMNZnk transcript — transcript markdown; source cache raw/sources/youtube-livestream-transcripts/htM02KMNZnk.txt (89,050 words).
• youtube q4Tr DknG2M transcript — transcript markdown; source cache raw/sources/youtube-transcripts/q4Tr-DknG2M.txt (4,039 words).
• youtube fL1iJHtl51Q — related YouTube source page.
• youtube fL1iJHtl51Q slides — slide evidence.
• youtube fL1iJHtl51Q reconstructed slides — slide evidence.
• youtube fL1iJHtl51Q dense slides — slide evidence.
• youtube htM02KMNZnk — related YouTube source page.
• youtube htM02KMNZnk slides — slide evidence.
• youtube htM02KMNZnk reconstructed slides — slide evidence.
• youtube htM02KMNZnk dense slides — slide evidence.
• youtube q4Tr DknG2M — related YouTube source page.
• youtube q4Tr DknG2M slides — slide evidence.

Novel Concepts / Clever Methods

• Agent-Ready Accessibility — Designing for agents and designing for accessibility converge around explicit structure, reachable controls, and understandable state.
• MCP Apps As Agentic App Runtime — MCP Apps treats interactive UI returned from MCP servers as a runtime layer for agent-facing software.

Evidence Boundary

This synthesis uses the official schedule plus cached video transcripts. Official AI Engineer World's Fair San Francisco 2026 livestreams and cut videos are primary event video sources for transcript/slide evidence; external, historical, or speaker-matched videos remain supporting context unless manually verified as exact official event recordings.

Official YouTube Recording

• youtube q4Tr DknG2M — official AI Engineer YouTube channel recording published 2026-07-15.
• Evidence status: youtube q4Tr DknG2M transcript; youtube q4Tr DknG2M slides.
• Boundary: use this recording as media evidence; keep date/time/room facts tied to the official schedule.