DeepLearning.AI Short Courses: Review

I’m Sofia, a CS student and serial free-course hunter. DeepLearning.AI’s short courses used to be my favorite loophole: real courses from the people who built the tools, certificates included, zero dollars. So the first thing you deserve to know is that the loophole shrank. As of August 2026, videos are still free — but labs, graded assignments, and completion certificates moved behind Pro, around $30/month or $25/month billed annually. If an older review told you “free including certificates,” it’s out of date. Check current pricing.

What’s still free is genuinely good. The catalog has 104 short courses (out of 127 total), each usually one to two hours, taught by Andrew Ng alongside engineers and founders from OpenAI, Anthropic, LangChain, Hugging Face, Meta, and 60+ other companies. The provenance is the whole point: ChatGPT Prompt Engineering for Developers packs nine lessons and seven code examples into ~1.5 hours and still tops Class Central’s 2026 ChatGPT list for API practice; Post-training of LLMs is taught by UW professor Banghua Zhu and covers SFT, DPO, and online RL — real frontier material in short-course form. Newer 2026 additions include a JetBrains AI-coding-workflows course and a non-technical prompting course. AI Python for Beginners claims 230,000+ learners with a built-in AI tutor.

My zero-budget strategy with this catalog: treat each course as a free evening briefing. Watch the videos, run what you can in the in-browser notebooks if your access allows, and don’t chase the certificate — chase the mental model. The videos alone keep me current on techniques I’d otherwise miss between semesters.

The caveats are real, and budget-hunters should hear them plainly. Content rot: the official forums are full of learners hitting errors because a course’s pinned API got deprecated — retired OpenAI model snapshots breaking notebooks, the Finetuning LLMs course’s lamini helper no longer working. Third-party learning paths warn that 2023–2024 courses largely use outdated APIs. With 100+ titles, many lag. Shallow-by-design coverage: reviewers consistently flag thin treatment of deployment, cost optimization, and evaluation discipline, and 46 courses on prompting alone can feel like quantity over curation. One more flag: a 2026 review roundup noted a Trustpilot complaint about unexpected Pro charges — read billing terms before any trial. And we found at least one older popular course URL that no longer resolves; treat the live catalog as truth.

Alternatives: the Hugging Face Agents Course competes directly and stays fully free, certificate included. For systematic depth, the Machine Learning Specialization or CS229 is the better spine. Nobody else gets you a LangChain briefing from Harrison Chase, though.

The catalog works best as a supplement to that spine. Pick a narrow gap, learn the vocabulary, then test it in a project before opening another short course.

Bottom line: keep this as your fastest free way to stay current — the videos cost nothing and the instructor roster is unmatched. Just go in clear-eyed: hands-on and certificates are Pro now, older courses may not run as recorded, and anything you want to stick needs deeper material after.