Manage tools and costs

What you will be able to do

Study AI without surprise bills and without risk to anybody’s data. Then turn everything from this module into an eight-week starter plan that you can begin next week.

Before you startWhat to know and to have readyAbout 35 minutes

What you need to know

• The five kinds of study time and the rule for each week, from Plan deliberate practice: Five kinds of study time and Plan deliberate practice: A rule for each week.
• The table of required courses, from Read the curriculum: All courses and their required courses.

What you need

• Your course order, under ## Course order in notes.md, from Read the curriculum: Guided practice: order a route.
• Your study week, under ## My study week, from Plan deliberate practice: Guided practice: plan your next week.
• Your domain card, under ## Domain card, from Domain knowledge in practice: Guided practice: write your domain card. The deliverable in your plan can use your domain.

Four kinds of practice environment

Environment	What it is	Cost	Use it for
Local exercises	Tools on your own computer: a text editor, a terminal, Python.	Free	Most practice in the foundation courses.
Synthetic data	Data made for practice. It looks real, but no real person or company is in it.	Free	Practice without privacy risk, with known correct answers.
Hosted notebooks	A notebook is a document that mixes text, code and the code’s results. Hosted notebook services, such as Google Colab and Kaggle, run it in your browser on their computers.	Free tier, with an account	Exercises that need more computing power than your computer has, or no installation.
Paid services	Language model APIs, cloud servers, and tools with paid plans.	You pay for use	Only when a course says so. Most courses have a free route.

The courses on this platform use the first three by default. Before any activity that needs an account or payment, the lesson says so, and offers a free route or a recorded result where it can.

Synthetic data: why the courses use it

The ticket data in the courses is synthetic. Larkfield is fictional, and so are its customers. This has three advantages:

1. No privacy risk. You can make mistakes, share your work and publish it, because no real person’s data is in it.
2. Known answers. The data was made with known problems, such as duplicates and wrong values, so you can check whether you found them.
3. The same data for everyone. Your results can be compared with the lesson’s results.

The limit: real data is messier than synthetic data. Later courses show you how to work with real public datasets too.

Rules for data that is not yours

Do not put data from your employer or your clients into a personal tool, a free online service or an AI assistant. The only exception is written permission from your employer. This includes customer lists, patient notes, contracts and financial records, even “just to test”.

If you want to practise with your domain, make your own synthetic version. Daniela, for example, writes 200 invented invoices with the same kinds of problems that she sees at work. She uses invented supplier names and amounts. She keeps her domain knowledge and leaves out the confidential data.

Keep costs under control

If a course uses a paid service, do these things before you start:

1. Set a spending limit in the service’s billing settings, where the service offers one, and a budget alert by email.
2. Write down every resource that you create: a server, a database, an API key.
3. Delete each resource at the end of the exercise. The course’s clean-up steps show how.
4. Check the bill one week later, to confirm that it is zero.

Free tiers change. A service that is free today may have limits or prices next year. Courses record the date when their instructions were last checked.

AI assistants while you learn

AI assistants can explain an error, give you another example, or ask you questions. They help most when you have already tried. They harm learning when they do the task for you, because then they practise and you do not. A good rule: try the task yourself first, and use an assistant to understand what went wrong. Then do the task again yourself.

Worked example: Daniela’s eight-week starter plan

Daniela’s route starts with Python and Mathematics and statistics, because both of her options need them. She studies 6 hours each week.

Week	Course and work	Hours	Kind of week
1	Python, Modules 1–2	6	Normal
2	Python, Module 3	6	Normal
3	Python, Module 4	6	Normal
4	Review Python, Modules 1–4. Repeat the hardest lesson.	3	Lighter
5	Python, Modules 5–6	6	Normal
6	Python final assignment	6	Normal
7	Mathematics and statistics, Modules 1–2	6	Normal
8	Deliverable: a Python script that finds duplicate invoices in 200 synthetic invoices, including near-duplicates such as INV-1043 and INV1043, with a short README.	6	Normal

Tools and costs: local Python and a text editor; synthetic data only. Cost: zero.

Dependency check: Python and Mathematics and statistics need no other course. The deliverable in week 8 uses only the Python course, which she finishes in week 6. No week needs knowledge that she has not learned yet.

The deliverable is small, but it is real: it uses her domain knowledge, it works, and she can show it. It is the first item in her portfolio, which you plan in Module 5.

Guided practice: your eight-week starter plan

1. In notes.md, add a heading: ## Eight-week plan.
2. Copy the template below.
3. Fill in eight weeks from your course order. Use your real study hours.
4. Make week 4 a lighter week for review.
5. Plan one small deliverable in week 7 or 8. It must use only what you learned before that week, and if possible, your domain.
6. Fill in the tools and costs line.
7. Do the dependency check: for each week, look at the table of required courses in Read the curriculum. Is every required course done in an earlier week, or already demonstrated?

Study hours per week:

| Week | Course and work | Hours | Kind of week |
|------|-----------------|-------|--------------|
| 1    |                 |       |              |
| 2    |                 |       |              |
| 3    |                 |       |              |
| 4    |                 |       | Lighter      |
| 5    |                 |       |              |
| 6    |                 |       |              |
| 7    |                 |       |              |
| 8    |                 |       |              |

Deliverable:
Tools and costs:
Dependency check:

Check: no week has more hours than you can really study. Every course in the plan comes after its required courses. The deliverable is something that another person could look at.

Common mistake

A plan that depends on what you have not learned

Symptom: in week 3 of your plan, you start the LLM applications course, but you have not done Python or LLMs. The first lesson asks you to write Python code that calls a model, and you cannot follow it.

Cause: the plan was written in the order that seemed most interesting, without the dependency check.

Fix: for each week, check the required courses in the table. Move the course to a later week, and put its missing prerequisites first. If you already have the skill from work, confirm it with the course’s final assignment first.

Check your understanding

Why does Daniela build her invoice script on 200 synthetic invoices, and not on real invoices from her job?

Show the answer

Real invoices belong to her employer and contain the data of suppliers and customers. She may not use them in personal tools or show them in a portfolio. Synthetic invoices keep her domain knowledge, the kinds of problems that she knows from work, without any confidential data. She can also publish them with her project.

Check your understanding

A course exercise asks you to create a cloud server. What do you do before you create it, and after you finish?

Show the answer

Before: set a spending limit or a budget alert, and write down the resource that you create. After: delete the server with the course’s clean-up steps, and check the bill a week later to confirm that the cost is zero.

Recap

• Practise locally, with synthetic data, and in free hosted notebooks. Use paid services only when a course says so, with a limit and a clean-up.
• Never put data from your employer or clients into personal tools. Make a synthetic version of your domain instead.
• Use AI assistants after you try, not instead of trying.
• Your eight-week plan has real hours, a lighter week, a dependency check and one small deliverable.

Next, in Module 5, you plan how to show your work so that employers trust it.

Next: Portfolio anatomy