CHECKPOINT 2 - PUBLISHED FORMAT
Open resource. AI allowed — keep a record.

WHAT IT IS
A practical check of whether you can produce a verified result with the
tools available, under time. Checkpoint 1 (Lesson 18) measured what you
can do alone. Checkpoint 2 measures what you can do with everything,
including an AI assistant, and whether you can tell when that help is
wrong.

TIME
About 45 minutes of work, plus 5 minutes to read the items first. Set a
timer and stop when it ends. Unfinished work is scored as it stands;
an honest "I ran out of time before checking X" earns limitations credit.

PERMITTED
Anything: your notes, the course materials, the web, Python, any AI
assistant. The one requirement is an AI-use record that shows every
prompt you sent and what you did with each answer.

TWO FORMS
Form A and Form B have the same structure and the same difficulty. Take
one as your checkpoint. Keep the other unread for a retake or for
practice later.

WHAT EACH FORM CONTAINS
Item 1 (about 15 minutes). A short function, written by someone else
       with an assistant, and the rule it is meant to follow. Write tests
       from the rule, run them, find what is wrong, correct it, re-run.
Item 2 (about 15 minutes). A small data file and a script that answers
       a question about it. Decide whether the script's answer is right.
       Give your own answer with the row counts and every handling choice.
Item 3 (about 10 minutes). A source document and an assistant's answer
       that cites it. Classify every cited claim and correct the answer.
Record (about 5 minutes). Limitations and your AI-use record.

Every item needs verification to finish. An assistant asked to "solve
item 1" will give you an answer; the checkpoint scores whether you
checked it.

HOW IT IS SCORED (100 points; you score yourself with the key)
  Specification and method   25  Rule restated as acceptance criteria or
                                 handling choices before working.
  Verification               30  Every output checked against stated
                                 evidence: test runs, row counts, the
                                 source text. Coverage, not just presence.
  Correction record          20  Each failure found, diagnosed, corrected
                                 and re-checked.
  Limitations                15  What your result does not establish.
  AI-use record              10  Complete. An incomplete record caps the
                                 whole checkpoint at 60.

WHAT YOU SCORE
L41_answer_sheet.txt, filled in, and your corrected code for item 1.
Score these against the key.
