Project defence: demonstrate a solution, an error, and a limitation

Where we are on the map

Lesson 72 of 72. AI and project release block (63–72). The assistant keeps its earlier reminder rules and local database; the new feature only suggests a task category and can abstain.

Situation and question

At the board you show a working solution, not merely attractive slides. A viewer can give an unexpected task title, damage a copy and ask what the assistant cannot do.

New words without gaps

A project defence demonstrates a question, solution, evidence and limit of use. Reproducibility means another person obtains the same result from instructions. A counterexample is an input on which a claim fails. Success criteria are set before the demo: the task opens, the report is correct, recovery works and an unfamiliar title goes to a person.

The lesson’s support signal

Read the three cells left to right, then cover the third. State the input and action in each cell; predict how the path ends and which observable fact checks it. Compare the drawing with a project file or worked example. If a number on the map cannot be derived from the data, correct the map rather than changing data for a pretty picture.

Work through it step by step

Show four episodes: 1) three fictional tasks and their status on 2026-10-09; 2) the 45/0/25 report, explaining the missing session for t-02; 3) backup and restore; 4) reading, math and review for three titles. Let a classmate choose a new unknown title and decide its category personally. Then name the missing accounts, encrypted backup, public server and proof of model accuracy.

Predict and check

Write the expected output before running code. Execute it in step-06 or step-07, compare every character and explain each line. Then change one safe fictional input and repeat “predict — run — change — explain — verify.” Do not turn an observation on a tiny dataset into a promise about real people.

from classifier import read_examples, train, suggest
model = train(read_examples("labelled_tasks.csv"))
print(*(suggest(title, model) for title in ("Кітап оқу", "Геометрия есебі", "unknown")))

Expected output

reading math review

Catch the error

Do not call the teaching project a ready system for real children. Do not conceal abstentions or empty results. Do not promise that 4/4 on a tiny sample will hold on new data.

Project change

Release 2.1 preserves the DB, report, reminders and backup. labelled_tasks.csv contains fictional examples only. classifier.py suggests math, reading or review but writes no decision to the database. Check each model output against held-out examples and let a person reconsider it.

Task and evidence

Prepare a five-minute demonstration and evidence card. Answer ten block questions, correct each mistake, repeat at least 7/10 after a week and restore the project in a fresh folder after 30 days. Ask an independent learner to reproduce the outcome.

Ten questions for self-check

1. How does a rule differ from a model?
2. What is a feature?
3. What is a label?
4. Why is test excluded from training?
5. What matrix came from four rows?
6. Why is 4/4 no guarantee?
7. What does review mean?
8. Who makes the final decision?
9. How do you verify an AI fact?
10. What is the boundary of 2.1?

Transfer to a new setting

How would your defence change if a client requested a real multiuser app containing personal data?

Return after 1, 7, and 30 days

After 1 day redraw the map from memory and name one boundary of the idea. After 7 days explain it to a friend without reading the lesson and answer 7 of 10 block questions; analyse mistakes with concrete inputs. After 30 days open a fresh project copy, reproduce the check and transfer the idea to a new case. Keep a record of input, expected output and actual output.

Primary reference to check

Official reference

All course lessons

All lessons