Online computer science courses: the best one is free, and it starts with C

The best online computer science course is free, and it is CS50x. That is an unusually easy recommendation to make, because Harvard published its introductory course as OpenCourseWare and nothing sold commercially improves on it for a beginner.

This page is about what computer science courses are for — as distinct from learning a language — and what the honest options cost.

Computer science is not the same as learning to code

Worth separating, because the two searches get answered interchangeably and they are different needs.

Learning to code is learning to make a machine do a thing: syntax, libraries, the tools around them. You can be genuinely productive with only this, and many working people are.

Computer science is what is underneath: how memory works, what an algorithm costs, why one approach takes a second and another takes an hour, what a compiler does to your code before the machine sees it. It is slower to learn, it feels less immediately useful, and it is what stops you being permanently mystified when something is slow.

If you want to build a website this month, do not start here. If you keep bouncing off explanations that assume you know what a pointer is, this is the gap.

CS50x — eleven weeks, free, and worth the C

Harvard's CS50x is free to work through via its OpenCourseWare, and its own site invites you to take it exactly that way. Eleven weeks of material, and the structure is the argument.

Weeks zero to four are C, which surprises people who came for Python. It is the point of the course. Writing C for a month means you have handled memory yourself, watched a program fall over because you asked for something you had not reserved, and understood what every later language is quietly doing for you. Then week five is data structures, week six is Python — which now feels like an enormous relief rather than a starting point — week seven is SQL, weeks eight and nine are the web, and week ten is a final project of your own.

The artificial-intelligence material sits in week seven alongside SQL, and there are separate CS50 courses for people who want to go further.

What the certificate is, exactly

Harvard is clearer about this than almost any commercial provider, and the sentence is worth reading twice:

Our free and verified certificates are not accredited academic offerings from Harvard or any of its affiliate institutions. We offer these certificates for self-educational and vocational purposes.

So: not a Harvard degree, not academic credit by default, not an accreditation. Some CS50 courses do appear in the American Council on Education's National Guide via edX's verified certificate programme, which is a separate mechanism with its own rules and is not the same as a university simply accepting it.

That precision is the standard this site holds everybody to. When a commercial platform calls its own product industry-recognised, it is worth noticing that Harvard declines to claim anything of the sort.

The deadline nobody mentions

Free does not mean open-ended. The CS50x FAQs ask what happens if you have not finished before end-of-day 30 June 2027: work carries forward if the course is extended, but work submitted in 2026 and earlier will not carry into a 2028 version, and if the course is updated you may only submit the new version's problem sets.

Practically, that means CS50x is a course you should start with an intention to finish rather than dip into for a year. Its problem sets are substantial and the final project is real work.

Do the paid platforms teach computer science?

Some of them offer something labelled that way, and it is worth knowing what it is. Codecademy lists Computer Science among the career paths where its professional certifications are available on Pro and All Access — its own exams, its own certificate. Its pricing, read on 24 September 2026, was $15.99 a month billed annually or $19.99 monthly for Pro.

That is a reasonable product. It is not a reason to pay, when the best course in the category is free and taught by the people who wrote the field's most-copied introduction. Pay for a platform because you need somebody to decide what you do next and to tell you why your code is wrong — not because a paid computer-science path sounds more serious than a free one. Every route and what it charges sets the options out with the dates we read them.

And if what you actually want is to build things rather than to understand them, the free Python routes are a better use of the same evenings.

What comes after CS50x, free

CS50x is an introduction, and the natural question on finishing it is what next. The honest answer is that the follow-ons are also free, and that you should pick by subject rather than by prestige.

CS50 runs further courses of its own — artificial intelligence with Python, web programming, databases — each free to work through on the same terms, each with the same certificate arrangement. Beyond that, the open material for algorithms and data structures from several universities is available at no cost, and the reason to do one is specific: algorithms are where computer science starts paying you back in ordinary work, because you stop guessing why something is slow.

What none of that requires is a subscription. The paid platforms cover computer-science topics, and cover them competently, but this is the one category where the free material is not a compromise in any dimension — it is the same lecturers, the same problem sets and the same certificate that Harvard declines to call accredited.