Reconstructed Slides: Training Agentic Reasoners — Will Brown, Prime Intellect

Source Video

Training Agentic Reasoners — Will Brown, Prime Intellect

Method

This deck is reconstructed from the existing video frame captures by detecting likely slide regions with OpenCV, cropping/upscaling those regions, deduplicating similar crops, and OCRing the cropped slide images locally. It is a cleaner companion to the full-stage frame deck.

Reconstructed Slides

• Source frame: slide-001.jpg
• Crop: full [0, 0, 960, 540] score 160.12

• Source frame: slide-002.jpg
• Crop: full [0, 0, 960, 540] score 175.8

• Source frame: slide-003.jpg
• Crop: full [0, 0, 960, 540] score 177.01

• Source frame: slide-004.jpg
• Crop: full [0, 0, 960, 540] score 177.12

• Source frame: slide-005.jpg
• Crop: full [0, 0, 960, 540] score 176.38

• Source frame: slide-006.jpg
• Crop: full [0, 0, 960, 540] score 176.38

Dense Scene-Detected Slide Candidates

• youtube PbHm2qKnu10 dense slides