Livestream Talk Segments
This page lists scheduled talks that were matched to a specific timestamp inside one of the broad AI Engineer World's Fair 2026 livestreams. Matches are generated from local timed YouTube captions and retained only when speaker names and talk-title terms agree in the same caption window.
Use these as navigational evidence into the livestream, not as a substitute for a cut talk video when a dedicated recording exists.
WF2026: Autoresearch & Keynotes (Day 2)
- Closing the Loop: An Autonomous AI Research Agent — 04:38:57 (score 92; matched: Tim Sweeney, research)
WF26: Harness Engineering & Startup Battlefield (Day 3)
- The Unreasonable Effectiveness of Separating the Task from the Model — 00:53:36 (score 120; matched: Maxime Rivest, effectiveness, model, separating, task, unreasonable)
- How Anthropic Builds: Lessons from Labs — 01:11:02 (score 99; matched: Mike Krieger, anthropic, labs)
- Thinner Agents on a Smarter Substrate: The Ontology-based Semantic Layer — 01:39:01 (score 106; matched: Emil Eifrem, based, ontology, semantic)
- Latent Space Live: the Inference Inflection from First Principles — 01:43:47 (score 106; matched: swyx, first, latent, space)
- Context Engineering in 2026: Compaction, Memory & Cost — 02:11:16 (score 92; matched: Louis-François Bouchard, engineering)
- Harness Engineering: Building the Production Cage for Powerful Domain Agents — 03:14:28 (score 99; matched: Mike Chambers, engineering, harness)
- Closing Keynote: Garry Tan — 07:57:20 (score 92; matched: Garry Tan, garry)
- $100,000 AIE Startup Battlefield — presented by Hyperagent — 08:18:26 (score 99; matched: Howie Liu, battlefield, hyperagent)
WF2026: Software Factories & Keynotes (Day 1)
- On AI and Knowledge — 00:10:28 (score 127; matched: Pablo Castro, knowledge)
- Recursive Model Improvement — 08:12:13 (score 141; matched: Lee Robinson, improvement, model, recursive)