Back to index
OtherSelf paced reference, a large library of primers and distilled notes you dip into rather than complete linearly·Free

Aman.ai (Aman Chadha)

4.4

Less a course and more a superb reference library. Aman.ai is where you go when you half understand transformers or attention or fine tuning and want a clear, dense, well organised primer to close the gap. It is free, it is deep, and for anyone already learning AI it is one of the most useful bookmarks you can have.

What We Liked

  • Genuinely deep, well written primers on exactly the topics people struggle with, from attention and transformers to LLM fine tuning and RLHF
  • Completely free with no gate, sign up or paywall
  • Excellent curation, including distilled Stanford course notes, paper summaries and reading lists that save hours of hunting
  • Kept unusually current for a free resource, with modern LLM and agent topics covered as they emerge

What Could Be Better

  • It is a reference, not a structured curriculum, so there is no path, no exercises and no hand holding
  • The density that makes it great also makes it a poor first introduction for a complete beginner
  • Quality varies a little across the huge topic spread, since some primers are more polished than others
  • You need enough grounding already to know what to look up and how to sanity check what you read

Detailed review

Aman.ai, built by Aman Chadha, is one of those resources that experienced learners quietly rely on far more than beginners realise it exists. The tagline about the art of artificial intelligence one concept at a time is a fair description of what it actually is, a very large and carefully organised set of primers that each take a single topic and explain it with real depth. The coverage is impressive and, importantly, it is weighted toward the things people genuinely find hard, so alongside the fundamentals of regression, neural networks and the maths behind them you get thorough, clearly diagrammed treatments of attention, the transformer architecture, tokenisation, fine tuning, reinforcement learning from human feedback, diffusion models, agents and the rest of the modern stack. What lifts it above a personal blog is the curation.

The site pulls together distilled notes for well known Stanford courses, summarises important papers, and maintains reading and resource lists, so a lot of the painful work of figuring out what to read and in what order has already been done for you. For a free resource it is also kept notably current, with newer language model and agent topics appearing while they still matter, which is not something you can say about most static references. The honest framing, and the thing to be clear about before you rely on it, is that this is a reference and not a course. There is no curriculum, no ordering, no exercises and nobody checking your understanding, so it does not hold your hand and it does not try to.

The same density that makes the primers so valuable to someone who is already learning makes them intimidating and easy to misread for a true beginner, who would be far better served starting with a structured course and coming here once they know enough to ask the right questions. Across such a wide spread the polish varies a little, with some primers reading like finished chapters and others closer to working notes, and because you are reading rather than doing you need enough grounding to sanity check what you take away. My take is that Aman.ai is one of the highest value free bookmarks in the whole field, provided you use it for what it is. Do not try to learn AI from a standing start here, but the moment you are in the middle of a proper course and hit a concept that will not settle, this is very often the clearest, most complete explanation you will find anywhere, and the fact that it costs nothing still slightly surprises me.

[ final ]

The verdict.

One of the best free reference resources in AI, but treat it as a companion rather than a starting point. Once you are past the absolute basics and actively learning modern machine learning, keep it open beside your course and reach for it every time a concept refuses to click.