Day 2: How AI Models Think
Open the lab the instructor calls out from the slide. Auto-saves to this browser — refresh-safe.
Three labs today. Your instructor will tell you which one to open and when. Each one auto-saves to this browser — refresh-safe.
Token-count three documents.
Three pre-loaded documents — a short customer email, a thread with quoted history, and a Letter of Credit. Estimate the token count first (characters ÷ 4), then verify in a counter, then log which window each fits. The middle doc is the trap.
Trace it. Diagnose it. Fix it.
A real LC-review output with its prompt + source. Tag every sentence by origin — prompt, training, or invented. Pick the worst error, diagnose it as lacks-vs-mistake, write the one-line fix.
Build your model-picker card.
One row per recurring task type from yesterday's triage list. Three levers — need, size, speed — land on a model class. Export the CSV before you close.