Markdown source

Dense Slides: WF2026: Autoresearch & Keynotes ft. Anthropic, Google DeepMind, Amazon AGI, Sonar, Arena, Recursive

Source Video

WF2026: Autoresearch & Keynotes ft. Anthropic, Google DeepMind, Amazon AGI, Sonar, Arena, Recursive

Method

This deck is slide-only. The existing captured video frame set supplies candidate frames, then local OpenCV rejects sponsor/title/speaker-only frames, crops visible slide surfaces, deduplicates, and saves the cropped slide images.

Cropped Visible Slides

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.96
• Text source: agent_vision.

Slide text:

CAPABILITY OVERHANG

Claude gets smarter in spiky ways

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.98
• Text source: agent_vision.

Slide text:

System Prompt Design

Small system prompt, few tools, lots of examples

Large system prompt, lots of examples, many tools

Smaller system prompt, tool search, no examples

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.97
• Text source: agent_vision.

Slide text:

Agenda

1. Meet OG Assist

2. The Origin Story

3. Betting on Effect

4. The Core Agent Loop

5. A2A, Evals & Sandboxing

6. Long Context Handling

7. Monitoring & Observability

8. Tools, Skills & Dev Workflows

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.95
• Text source: agent_vision.

Slide text:

Origin Story

One bet on agents, one immediate yes.

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.94
• Text source: agent_vision.
• OCR decision: ready — Code and small explanatory text are better handled by OCR than manual transcription.

Slide text:

Our agent loop, rebuilt Effect-native.

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.95
• Text source: agent_vision.
• OCR decision: ready — Dense diagram labels and small body text make OCR preferable.

Slide text:

Production Evals for Agentic Systems

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.91
• Text source: none.
• OCR decision: ready — Cropped headline plus many small chart labels are OCR-suitable.
• Slide text: not surfaced (illegible by AI classifier).

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.95
• Text source: agent_vision.
• OCR decision: ready — Dense diagram slide with small subtitle/body text is OCR-suitable.

Slide text:

Production Evals for Agentic Systems

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.98
• Text source: agent_vision.
• OCR decision: ready — Diagram slide with multiple small labels and a central illustration; OCR will capture labels more reliably than manual reading.

Slide text:

The Anatomy Of Agentic Failure

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.99
• Text source: none.
• OCR decision: ready — Dense diagram with several small callout boxes and circular annotations; OCR is the right capture method.

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.76
• Text source: agent_vision.
• OCR decision: ready — Large embedded product UI screenshot with many small labels and UI chrome; OCR will be more reliable than hand transcription.

Slide text:

Engineering the future of AI

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.84
• Text source: none.
• OCR decision: ready — Product UI screenshot dominates the frame and contains many small interface labels; OCR is suitable.

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.97
• Text source: agent_vision.

Slide text:

Principle 5 Modularity

Which capabilities should be reusable, and which stay local?

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.96
• Text source: agent_vision.

Slide text:

The agent is its data

Specifically: the log

Classification audit: raw/sources/slide-ai-classification/dense/4sX_He5c4sI/audit.json