Learn AI, ProperlybyLinksley.com

Public board

A route from using AI to understanding it — free courses, primary sources, and where practitioners actually argue.

Use it, then understand it, then build with it

Foundations → hands on → primary sources → stay current

Most people skip straight to building agents and then cannot tell why one prompt works and another does not. The order below is deliberate: enough foundation to reason about what the model is doing, then practice, then the papers. Everything in the first three groups is free. The fourth is where the field argues in public, which is where you learn what is actually contested.

Foundations

1DeepLearning.AI short coursesFree, an hour each, built with the companies whose tools you are using.2fast.aiTop-down: build something that works in lesson one, understand it by lesson six.33Blue1Brown — neural networksThe clearest visual explanation of what is happening inside. Watch before any maths.4Hugging Face courseFree, practical, and ends with you having actually trained something.

Primary sources

9arXiv cs.LGWhere it appears first. Read abstracts; skim the rest without guilt.10Papers with CodeThe paper and the implementation together.11DistillArchived, not dead — still the best-explained machine learning on the internet.12Anthropic researchInterpretability and safety work, written to be read.

A route through, in order

0/6

Watch the 3Blue1Brown series — the whole thing, once
Finish one DeepLearning.AI short course end to end
Run a model locally with Ollama and notice what changes with size
Complete one Kaggle notebook without copying the solution first
Read one paper properly, including the limitations section
Build one small thing nobody asked for and write down what broke

Hands on

5Google ColabA free GPU and a notebook. No setup between you and an experiment.6KaggleReal datasets, public solutions to read, and competitions to measure yourself against.7Hugging Face SpacesThousands of running demos with their source visible.8OllamaRun a model on your own machine — the fastest way to build real intuition about size and speed.

Stay current

13Simon Willison's blogThe most reliable practitioner running commentary on what actually shipped.14Import AIWeekly, technical, and unusually good at separating signal from launch noise.15Hacker NewsThe comments are the point — that is where claims get tested.16LMArenaWhat people prefer in blind comparison, updated continuously.

Who this is for

People moving from using AI to understanding it

No bootcamp, no paid certificate. The route works if you do it in order — the temptation is to skip to building agents, and that is exactly where people get stuck without knowing why.

The thinking behind this boardFive courses, five portals, and finals in a week

Made with Linksley

Your links could live on a page like this.

Boards are shared bookmarks with real permissions — for a team, a family, or a page you publish like this one. Free on one board, no card.

Make your own board Browse more boards