From Neural Networks to Agents — Slides | Original
Planning to give a talk titled “From Neural Networks to Agents” in an engineering group at my company in the next few weeks. Wrote these slides to walk through:
- Neural networks from scratch — MNIST, backprop, dimensions
- Transformers & GPT — Q/K/V, attention, multi-head
- nanoGPT training on H200 — FineWeb, I/O bottlenecks
- Claude Code tips — precise input, paste-don’t-describe
- Agents — OpenClaw & Hermes — exec as first-class, tool calls at machine speed
Slides (Marp): PDF · Preview PDF