Course overview¶

Python is not the subject. The subject is how to make a computer answer a scientific question without quietly changing the question.

The recurring loop is:

question → representation → transformation → verification → evidence

Syntax matters in the same way notation matters in mathematics: it must be precise, but memorising notation is not the same as knowing what to do.

Who this is for¶

The course assumes:

comfort with algebra, graphs, and physical units;
no prior programming after completing the Python preflight, or equivalent familiarity;
willingness to predict an answer before seeing a plot.

It also works as a reset for students who can already make notebooks run but have not yet built a tested script, command-line tool, or reproducible figure.

What you will be able to do¶

By the end, you should be able to:

translate a physical question into explicit inputs, outputs, and failure modes;
trace values, shapes, units, and state through unfamiliar code;
design small functions with testable scientific contracts;
use NumPy arrays without losing track of axes or meaning;
debug by shrinking a failure and testing a hypothesis;
validate tabular data before reducing it;
design a plot around a claim rather than around default settings;
test limiting behaviour, conservation, and numerical convergence;
turn an analysis into a path-safe command-line tool;
build a locked, one-command reproduction capsule.

The lesson rhythm¶

Every lesson has the same five moves.

Frame. What decision are we making? What could make it wrong?

Predict. State shapes, units, limits, or trends before execution.

Build. Express the smallest transformation that can decide.

Verify. Attack it with a claim that could fail.

Reflect. Say what the result supports and what it does not.

Quick route¶

Allow 6–8 focused hours. Do these in order:

Then inspect the reproduction capsule. If any part feels like magic, return to the linked full-route lesson.

Full route¶

Allow 20–24 hours including exercises.

Stage	Lessons	Scientific habit
Model	01–04	make state and logic explicit
Interrogate	05–07	fail loudly, preserve schemas, argue with plots
Trust	08–10	verify numerics and leave a reproducible trail
Practise	11–12	turn CoMPhy-style output into validated evidence
Integrate	Labs 12–13	join reductions, figures, commands, and checksums
Publish	Capstone 13	build one-command public science

Do not binge-read the text. A useful cadence is 20–30 minutes of reading, 45–75 minutes of exercise, and 10 minutes explaining your result to another person.

What is deliberately not here¶

There is no early catalogue of every string method, collection type, class feature, or plotting option. You will meet language features when a problem creates a reason for them.

There is also no notebook-only route. Notebooks are excellent scratch benches. They are poor final contracts when execution order, hidden state, and large outputs become part of the result. Canonical work in this course ends as functions, scripts, tests, and small data products.

Assessment¶

The capstone is a small Basilisk-to-publication reproduction capsule. It needs:

one physical question and nondimensional control parameter;
one public or supplied dataset;
a command-line tool that validates inputs;
reduced data and one scientific figure;
tests for schema, limiting behaviour, and a numerical regression;
a locked environment, provenance note, and one-command rebuild.

The standard is not code cleverness. The standard is whether another student can see, challenge, and reproduce the chain of reasoning.