Rules, algorithms, and trainable models

Where we are on the map

Lesson 63 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

The assistant already applies an exact rule to remind us of due dates. Now it may suggest whether a new task concerns reading or maths. The due-date rule stays unchanged: a category suggestion never controls reminders.

Where to get the project files

Open the checkpoint folder, download the repository with Code → Download ZIP and find course/informatics-assistant/step-07. Open a terminal there. python3 --version shows Python; then run the lesson check. All tasks are fictional; enter no real personal data.

New words without gaps

An algorithm is a finite sequence of steps. A rule is specified by a person in advance, such as “remind when at most two days remain.” A trainable model derives parameters from examples; here they are word counts. A prediction is a suggestion for a new record, not proof. review means the evidence cannot distinguish categories.

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

Compare two paths on one card. For t-01, a precise due date yields REMIND under the 1.0 rule. The title Кітап оқу may yield reading through the model’s word counts. Change the title to unknown words: the model should return review while the due-date rule still works. Model quality and reminder correctness need separate checks.

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 datetime import date
from assistant_core import reminder_status
from classifier import read_examples, train, suggest
rows = read_examples("labelled_tasks.csv")
print(reminder_status({"done": False, "due_date": "2026-10-10"}, date(2026,10,9)))
print(suggest("Кітап оқу", train(rows)))

Expected output

REMIND
reading

Catch the error

Code does not become AI merely by using if. A trained model does not understand a word as a person does. Do not replace an exact due-date rule with a statistical guess.

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

Draw two tracks for one task: due date → status and title → suggested category. Name the input, output and person responsible for the final choice.

Transfer to a new setting

A thermostat starts heating below a chosen temperature, while a system forecasts tomorrow’s weather. Which path uses a fixed threshold?

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

Next lesson

Features, labels, and an honest dataset