Is a paid course worth it when the vendors publish free guides?

Only if it adds structure, exercises and evaluation. The free documentation is excellent for capabilities and useless for judgement. Pay for practice and critique, not for a walkthrough of features.

Does prompt engineering still matter?

Less as a bag of tricks, more as clear specification. Newer models need less coaxing and still need to be told what good looks like, what to avoid and what format to produce. That is writing a brief, and it does not go out of date.

Which model or tool should a course use?

Whichever your workplace already pays for. The transferable part is the method, and a course tied to a tool you cannot use turns every exercise into a translation problem.

Are there credentials worth having?

A few fundamentals exams cover the vocabulary and cost little: the AWS AI Practitioner and the Azure AI Fundamentals exam from Microsoft are the two with pages of their own. Nobody senior is hired on them. Demonstrated use inside a real job is the credential this field reads.

What is retrieval-augmented generation and do I need it?

Grounding a model in your own documents so its answers cite sources rather than memory. You need it the moment the questions are about your data rather than the world's, which for most workplace use is immediately. The concept page explains the mechanism; the builder courses here teach the plumbing.

Generative AI & ChatGPT Courses

Work with the new generation of AI tools - prompt engineering, ChatGPT workflows, image generation, and building on top of LLMs

This is the fastest-decaying section in the directory. A course recorded eighteen months ago may show menus that no longer exist, quote limits that have doubled, and demonstrate workarounds for problems the models no longer have. Yet underneath the churn there is a durable skill, and telling the two apart before you pay is the whole art of choosing here.

Durable underneath, disposable on top

Lasts	Does not last
Knowing what these systems do well enough to predict where they fail	The exact wording of a prompt that worked once
Writing an instruction that constrains the output instead of hoping for it	A list of the best tools this quarter
Grounding a model in your own documents so it answers from sources	The position of a button in an interface
Evaluating whether the output is right rather than whether it reads well	Workarounds for limits the next release removed

A course that is mostly the right-hand column is a magazine article with a price. The left-hand column is what the AI concepts are about: prompt engineering as specification rather than incantation, retrieval-augmented generation for grounding, and large language models as a thing with knowable failure modes. Read those before buying, and the disposable courses become easy to spot.

Choosing a course in a field that moves monthly

Check when the material was last updated, not when the listing was refreshed. Screenshots date a course faster than its description does.
Prefer courses organised around problems you have, rather than around a tool's feature list.
Treat any promise of a repeatable formula with suspicion. If a single prompt template worked reliably, it would be a product rather than a course.
Evaluation content is the strongest quality signal there is here. Very few courses teach how to tell whether the output is good, and it is the skill employers are short of.

Two audiences, two different categories

Most people here want to work better: drafting, summarising, analysing, automating the parts of a job that were always tedious. That needs judgement and practice, not engineering, and the return arrives within weeks. A smaller group wants to build: retrieval over private data, agents that take actions, evaluation harnesses, cost and latency budgets. That needs programming and an understanding of the model layer, and it belongs beside the AI concepts and the AI portal rather than beside a tools tour. Courses rarely say which audience they are for, so read the exercises rather than the headline.

Before you paste anything

Large Language Models

Large language models are neural networks trained on a vast quantity of written text to predict the piece of text that comes next, an objective narrow enough to state in a line and broad enough to yield writing, translation, code and summary as side effects.

The field, explained: Artificial Intelligence & Machine Learning

This category belongs to the field Data, Analytics and AI, which maps the topics, concepts and certifications behind these courses.

Frequently asked

Is a paid course worth it when the vendors publish free guides?: Only if it adds structure, exercises and evaluation. The free documentation is excellent for capabilities and useless for judgement. Pay for practice and critique, not for a walkthrough of features.
Does prompt engineering still matter?: Less as a bag of tricks, more as clear specification. Newer models need less coaxing and still need to be told what good looks like, what to avoid and what format to produce. That is writing a brief, and it does not go out of date.
Which model or tool should a course use?: Whichever your workplace already pays for. The transferable part is the method, and a course tied to a tool you cannot use turns every exercise into a translation problem.
Are there credentials worth having?: A few fundamentals exams cover the vocabulary and cost little: the AWS AI Practitioner and the Azure AI Fundamentals exam from Microsoft are the two with pages of their own. Nobody senior is hired on them. Demonstrated use inside a real job is the credential this field reads.
What is retrieval-augmented generation and do I need it?: Grounding a model in your own documents so its answers cite sources rather than memory. You need it the moment the questions are about your data rather than the world's, which for most workplace use is immediately. The concept page explains the mechanism; the builder courses here teach the plumbing.

Why Learn Generative AI & ChatGPT?

The fastest-moving corner of the catalog: using ChatGPT and its rivals well, prompt engineering that goes beyond party tricks, image generation, and developer-side work with LLM APIs, agents, and retrieval pipelines. For professionals folding AI into their work and builders shipping AI products.

Last reviewed 26 September 2026 · Getting Digital