CHECKPOINT 1: CLOSED BOOK  -  BLUEPRINT
Checkpoint 1 covers Units I and II (lessons 1 to 17). It is a self-check,
not an exam, and this checkpoint is not scored. You work a set closed
book, then open the answer key and mark yourself against the standard
it gives.

FORMAT
  Time:        About 45 minutes per set. An estimate, not a limit. Note
               how long you took.
  Resources:   None. Closed book. No AI, no notes, no reference sheet,
               no calculator, no network. Do not run any code until you
               have finished the set.
  Answer on:   Paper, or a blank text file. Writing the answer out in
               full is part of the measurement.
  Sets:        Two parallel sets built to this blueprint, Set A and
               Set B. Do one now and the other a week or two later.
  Key:         One answer key per set, at the end of the set. Open it
               only after you finish.

WHY CLOSED BOOK
  Nobody will check, so the condition protects only you. The checkpoint
  exists to find out what you can do unaided, because that is what you
  fall back on when an assistant is wrong, unavailable or not allowed,
  and it is the reading skill you rely on whenever you sign for
  generated code. A wrong answer you got on your own tells you exactly
  which lesson to reread. An answer you looked up tells you nothing.

CATEGORIES
  Mechanism                  3 items
      Short written answers. Next-token prediction, sampling and
      temperature, training versus inference (weights fixed,
      conversation is context), the three circles.
  Code reading and tracing   5 items
      Predict exact output; state which branch runs; complete a trace
      table for a loop; state a loop's stopping condition; read a
      function's signature and body.
  Fault diagnosis            2 items
      Given code and its real output: name the line that raised, state
      the cause in plain words, find the line that holds the fault,
      judge a proposed fix. One program crashes; the other runs cleanly
      and prints a wrong answer.
  Method                     1 item
      Decompose a short request into function-sized sub-problems,
      surface open assumptions, write a signature and testable
      acceptance criteria with a boundary.
  Verification reasoning     1 item
      Given an AI output, list the checkable claims, name the authority
      and evidence that would settle each, and predict the result where
      the code is shown.

ITEM TYPES
  - Short written answer (2 to 6 sentences). Judged on whether the
    mechanism you describe is right, not on vocabulary.
  - Predict the exact output of a short program (about 10 lines).
  - Trace table: the value of a variable after each pass of a loop.
  - Traceback reading: the program and its full error output are given.
  - Structured written problem: decomposition, assumptions, signature,
    acceptance criteria.
  Show your reasoning. If you slip, the written reasoning shows you
  where, and in a trace table it lets you count only the first wrong
  row.

WHAT THE CODE USES
  Only what L10 to L17 taught: variables; int, float, str and bool; +,
  -, *, /; comparisons (<, <=, >, >=, ==); if / elif / else; for loops
  over range() or over a list; while loops; len(); int() and str();
  print(); lists; functions with def and return. The programs are new
  to you.

WHAT IS NOT ON IT
  - No questions about how to use any particular AI tool, product or
    menu, and no prompt-writing tricks.
  - No model architecture (attention, layers, transformers), no maths
    of training.
  - No memorised facts: no context-window sizes, model names, release
    dates or statistics.
  - No writing a whole program from scratch, and no syntax you were not
    taught (no list comprehensions, dictionaries, classes, file
    handling or imports).
  - No hardware, networking or encoding questions from L2 to L4, except
    as they bear on tokens.
  - No trick questions. Every program is plain and does what its text
    says; the question is whether you can read it.

HOW TO PREPARE
  The skills here are the ones you used in the lessons with no tool:
  predict before you run, read a traceback from the bottom up, explain
  the mechanism in your own words, write criteria with a specific input
  and expected result. The practice set in the L18 materials is in the
  same style as the two sets, with its own answer key.
