Guide · updated November 1, 2026
Connect the pieces
The standing explanation. Today’s briefing is what changed. This is how the pieces fit, what they cost, and how a team gets better at one job.
Begin with the field card (three questions, prices, worked checks). Then ranks and prices, agents and loops, and fine-tuning for one job. Today’s briefing is what changed; these chapters stay.
Field card first
How to read an AI claim
Three questions, three worked checks from the briefing, and four list prices — then the long chapters below.
Then · chapter 1 of 3
How models are ranked, and what they cost
Standard list prices per million tokens — GPT-6.1 Terra $1/$5, Claude Opus 5.5 $4/$20, Gemini 4 Argon $2/$10 — and how to read a rank. Leave with three habits: name the test, name the price, name what is still unknown.
Then · chapter 2 of 3
Agents, loops, and tools
A model writes the next tokens. The harness around it — tools, memory, context, and a check — decides whether the job finishes. Leave asking which piece moved: model, tools, loop, or the check.
Then · chapter 3 of 3
Fine-tuning for one job
Fine-tuning changes a model you already have, using examples of a decision your team can score. It is how a company owns a narrow task without training a foundation model. Leave with one scored decision, a held-out set, and a comparison to the model you pay for now.
Model evaluation — a number belongs to the test, the date, and who measured it.