391 free AI courses, checked: a searchable catalog with link health, provenance, and 2026 updates
A viral "replace your degree" list of free university AI courses. Every link checked, every course searchable, with credit to where the list really comes from, an honest look at its age, and the modern courses (including robotics) it's missing.
Start here (no maths needed)
A giant list of free university AI classes. Lots of it is gold, some links are broken, and the newest class is from 2022.
Think of it like this
Imagine someone hands you a huge library card catalogue from a few years ago. Most books are still on the shelves, but some have gone missing, and none of the newest books are listed. This page checks which "books" are still there, lets you search them, and adds the new books the old catalogue missed.
How it works
- Every course from the list is in the searchable table below.
- Each link is coloured: green works, red is broken, amber needs a manual check.
- Pick a topic (for example "Reinforcement Learning") and tick "working links only".
- For newer topics like ChatGPT-style models and robots, use the "2026 updates" section.
Why robots care
The list has reinforcement learning and computer vision, which robots need. But it has no robotics section at all, so this page adds free robotics courses (robot arms, walking robots, learning from demonstrations).
Try it
In the catalog, choose Category: Reinforcement Learning, set From year to 2019, and tick Only courses with a working link. How many are left?
Words to know
- Curriculum
- A planned list of what to learn, in order.
- Link rot
- When old web links stop working.
- Provenance
- Where something really came from.
- Lecture
- One recorded class.
Going deeper: provenance, the reality check, the catalog, and the 2026 updates.
Track: LLM + Robotics. Mostly foundations (deep learning, ML, RL, computer vision, NLP) that feed both tracks. It's missing post-2022 LLM material and robotics, and the updates section covers both.
1. Summary
TL;DR
- What it is: 391 free courses in 14 categories, from deep learning (81) and general ML (63) to summer schools (58) and computer vision (49).
- Link health: I checked all 733 unique links. 610 load, and of the 328 YouTube videos and playlists, 306 still exist. 381 of 391 courses still have at least one working link.
- Age: the newest course is from 2022, and most date from 2017–2021. Nothing covers modern LLM training, post-training, serving or robot foundation models.
- Provenance: the article reproduces the Deep Learning Drizzle GitHub list without credit.
| Category | Courses | Category | Courses |
|---|---|---|---|
| Deep Learning | 81 | Graph Neural Networks | 17 |
| General Machine Learning | 63 | NLP | 16 |
| Boot Camps / Summer Schools | 58 | Medical Imaging | 9 |
| Modern Computer Vision | 49 | Probabilistic Graphical Models | 8 |
| Reinforcement Learning | 28 | Automatic Speech Recognition | 6 |
| ML Fundamentals | 24 | Bayesian Deep Learning | 5 |
| Optimization for ML | 23 | Bird's-eye view of AGI | 4 |
2. Provenance: where the list comes from
The article's content matches Deep Learning Drizzle, a GitHub list by kmario23 (created Nov 2018, about 13,000 stars, still updated in 2026). The evidence is in the article itself:
- Its "Go to Contents ⤴️" links point to
github.com/kmario23/deep-learning-drizzle#contents. - It ends with the repo's own "To-Do" list, including unfinished sections on autonomous vehicles and computer graphics.
The article doesn't credit the repo. Credit belongs to Deep Learning Drizzle's maintainer and to the hundreds of instructors who made their courses free. The repo has no licence file, so this module links to it for the canonical, maintained version. The catalog below is a personal, searchable index of the course names and links, with link-health checks added.
3. Reality check
- "Replaces your degree": free lectures can teach the material, but a degree also brings structure, assessment, feedback, peers and a credential. Courses with assignments (CS231n, CS224n, CS285, CS336) come closest. Watching videos alone doesn't.
- "5000+ lectures": that's plausible given 391 courses, but quantity isn't the point. Nobody needs 5,000 lectures. A focused path of 5–10 courses with exercises beats it.
- Freshness: the list stops at 2022. The foundations (backprop, CNNs, RL theory, probabilistic models) haven't changed. LLM practice has changed completely since 2022, and so has robot learning (VLAs, diffusion policies).
- Link rot is real but modest: about 1 in 8 links are dead, and 10 courses have no working link left.
4. The catalog
Interactive: search all 391 courses
● works · ● dead (404, timeout, removed video) · ● check manually (blocked to scripts, or a private/non-embeddable YouTube video)
| Course | Year(s) | Who | Links |
|---|
5. 2026 updates: what to take instead or as well
Keep the classics for foundations, and add these for what's changed since 2022. Every link below was checked.
| Category in the list | Still great from the list | Add for 2026 |
|---|---|---|
| Deep Learning | Use the catalog's Stanford/Berkeley/CMU entries with assignments | Karpathy, Zero to Hero · fast.ai · Module 06 (30 papers) |
| NLP | Stanford CS224n | Stanford CS336 · Hugging Face LLM Course · Module 11 |
| Reinforcement Learning | Classic RL theory courses in the catalog | Berkeley CS 185/285 Deep RL · Hugging Face Deep RL Course · RLHF Book · Module 12 · GRPO |
| Computer Vision | Stanford CS231n notes · CMU 16-385 | Module 08 · multimodal & VLAs |
| Graph Neural Networks | The catalog's GNN entries | Stanford CS224W · Module 06 · Day 21 |
| Optimization | The convex optimisation courses in the catalog | Module 12 · AdamW · Muon |
| Missing: LLM systems & inference | None | Module 09 · 16 · MIT 6.5940 TinyML & Efficient DL · Module 08 course map |
| Missing: post-training & alignment | None | Module 07 · DeepLearning.AI, Post-training of LLMs |
6. The missing robotics section
The original list's own To-Do includes an "Autonomous Vehicles" section that was never added, and there's no robotics category at all. Free courses to fill the gap (all links checked):
| Course | What it covers | Fills which gap |
|---|---|---|
| Modern Robotics (Northwestern) · Coursera specialisation | Rigid-body motion, forward/inverse kinematics, dynamics, trajectory generation, control | Kinematics and control, flagged as gaps in Modules 02 and 14 |
| MIT Robotic Manipulation (Russ Tedrake) | Perception, planning and control for robot arms, with notes and code | Manipulation end to end |
| MIT Underactuated Robotics (Russ Tedrake) | Dynamics, optimal control, trajectory optimisation, legged locomotion | Control and MPC |
| Berkeley CS 185/285 Deep RL | Deep RL with a strong robotics flavour | RL for robots |
| Hugging Face Robotics Course + LeRobot | From classical robotics to learned policies on real, low-cost hardware | Imitation learning and VLAs, hands-on |
| MuJoCo · Isaac Lab | Physics simulators for robot learning (docs and tutorials) | Simulation and sim-to-real |
7. Self-check
Why credit Deep Learning Drizzle rather than the article?
The article's own internal links and its To-Do list come from that repo. The list was compiled and maintained there since 2018.
What fraction of courses still have at least one working link?
381 of 391 (about 97%), even though about 1 in 8 individual links are dead.
Which major topics are missing because the list stops in 2022?
Modern LLM training at scale, post-training and alignment, LLM serving and inference, agents, and robot foundation models (VLAs, diffusion policies). There's also no robotics section at all.
What makes a free course closest to "replacing a degree"?
Graded assignments and projects you actually do, in a sequence, ideally with peers or feedback. Watching lectures isn't enough.