WF2026: Autoresearch & Keynotes ft. Anthropic, Google DeepMind, Amazon AGI, Sonar, Arena, Recursive
summary
Roll the tape if concrete claims matter: Anthropic pitches the new Fable mythos class of models with code‑execution overhangs not vague demos; Sonar shows a zero‑trust multi‑layered verification workflow that customers report cuts AI‑derived production outages ~44% and a bank saw ~92% fewer issues when verification is baked in; Amazon open‑sources perception agent tools (on‑page annotation + visual verification) that run HTML reports and self‑check changes; Recursive’s auto‑research finds novel kernels and cuts nanoGPT bits‑per‑byte to 0.91.
AI-written summary · about
chapters (20)
- 0:00AI Engineer Worlds Fair 2026
- 11:14Conference opening
- 16:53Fable models intro
- 35:45Sonar code verification
- 54:35Amazon AGI workflows
- 1:16:10DeepMind code engineering
- 1:36:55Evals track
- 1:43:30OG Assist system
- 2:00:56Eureka Machine
- 2:16:25Meta research & evals
- 2:40:56Nano Banana 2 APIs
- 3:00:56AI research OS
- 3:40:05Arya agent demo
- 4:34:39Character AI personas
- 5:00:56Parameter Golf results
- 5:22:04Agent relocation systems
- 5:56:20AI cost vs intelligence
- 7:16:27Nori AI coding agents
- 7:31:45Mobile AI workflows
- 7:44:36Closing remarks
related talks
26mAl and Software: The Good, The Bad, The Scary
01h 01mKeynote: Guardians of the AI Era: Navigating the Cybersecurity Landscape of Tomorrow
07h 54mAIE Miami Keynote & Talks ft. OpenCode. Google Deepmind, OpenAI, and more!
13mMeet the Speakers
27mBuilding Generative AI features
07h 44mAI Agents Conference 2026 - Day 1
09h 28mDay 2
30mKeynote: Fireside Chat