TokenmaxxingAI token costsAgents online
GuidesLabsEventsLeaderboardToolsBriefing
Events / Washington, DC

Build a Reasoning Model (from Scratch)

Hands-on study group working through Chapter 5 of a reasoning-models book, covering inference-time scaling via self-refinement -- improving model reasoning without retraining -- with attendees coding along from a shared repo.

Tue, Aug 4, 11:00 PMHerndon Fortnightly Library Conference Room · Herndon · VA
Event page on MeetupMore in Washington, DC

Why it matters

Inference-time scaling is precisely the tradeoff tokenmaxxing readers weigh constantly: spend more tokens per query for better reasoning, or route to a cheaper single-pass call -- this group works the mechanics hands-on.

The tokenmaxxing angle

Self-refinement is a token-cost lever by definition -- it burns extra inference tokens per request in exchange for answer quality, the exact knob practitioners tune when deciding how much spend a given query deserves.

From the organizers

Session covers Chapter 5, on inference-time scaling via self-refinement, with attendees encouraged to pull code beforehand from the GitHub repo rasbt/reasoning-from-scratch to follow along.