I stopped trusting AI headlines, so I built a course
I kept seeing the same kind of AI system described as a miracle in one headline and a failure in the next. One story gave the model a mind. The next called it useless. Both wanted a reaction before they explained the mechanism.
I got tired of borrowing my opinion from headlines. So I went backward.
I followed the short chain of papers behind the tools: early neural language models, word vectors, sequence-to-sequence translation, attention, the transformer, scaling laws, and training from human feedback. Then I moved into the papers and documentation on hallucinations, retrieval, long context, evaluation, infrastructure, and the systems that give a model data and tools.
There was less magic than the headlines promised. The engineering was more interesting.
The split that made the noise quieter
My notes kept returning to a simple separation:
- The model predicts text.
- The surrounding system supplies documents, memory, tools, and permissions.
- Evaluation tells you whether the result works on the real task.
Once those layers were clear, AI news stopped reading like a thriller. It became an operations update. What changed in the model? What changed in the surrounding system? What was measured, and what still needs verification?
That is also a good way to avoid drowning in daily noise. You do not need twenty channels repeating the same launch. You need a stable map, primary sources, and enough practical work to recognize which layer actually changed.
Those notes became AI Academy
I turned the map into a free, self-contained course. AI Academy has 5 connected courses, 29 modules, 103 lessons, 213 knowledge checks, 8 visual study maps, 5 applied cases, and 172 primary sources.
The route starts with what a language model does. Then it moves into workplace adoption, data sources and RAG, infrastructure, and a hands-on lab where the pieces become a working system. You can read it in order or choose the path that fits your work.
The course is available in American English and Brazilian Portuguese. It works without an account and keeps progress in your browser.
The course content and original diagrams use the CC BY 4.0 license. The code uses MIT. Take it, study it, adapt it, and keep the attribution.
You do not have to memorize every paper. You need to know which layer changed when the next loud headline arrives.