Why it matters
Practitioners who understand how reasoning models are constructed internally make better decisions about when to use chain-of-thought, extended thinking budgets, and multi-step inference — all of which have direct cost implications at scale.
The tokenmaxxing angle
Knowing the internals of reasoning models helps engineers decide when to pay the premium for o-series or Claude extended-thinking versus routing to a cheaper non-reasoning model. The coding exercises use the rasbt/reasoning-from-scratch GitHub repo as the hands-on reference.
From the organizers
The session works through two chapters — 'Understanding Reasoning Models' and 'Generating Text with an LLM' — with code at github.com/rasbt/reasoning-from-scratch. Held at University of North America, 12750 Fair Lakes Cir, Fairfax, VA.