everything-ai-ml: a curated AI/ML resource index that lives in a TypeScript file

The gap everything-ai-ml fills: link rot in AI reading lists

Most AI reading lists live in a pinned social post or a note that nobody updates. everything-ai-ml takes the opposite route: it keeps the list in a repository, splits it into roughly thirty themed sections, and publishes it two ways, as a README and as an interactive cheatsheet at viveknaskar.github.io/everything-ai-ml. The audience is narrow and clear. It is for someone who already knows they want to learn about RAG or MLOps and needs a starting set of links, not for someone who wants a course with exercises. The section list in the README runs from AI/ML Key Concepts and Building Blocks through Prompt Engineering, RAG, Fine-tuning, Agentic AI, MLOps and GenAIOps, Security, Google Cloud AI and ML, AI Cost Optimization, Quantum Computing and PQC, Courses, Certifications, Books, Must-Read Research Papers, YouTube Channels, Research Blogs, Communities, Practice Problems and Interview Preparation. That is a breadth-first index. Each entry is a title, a URL and a one-line description, for example Supervised Learning pointing at a Medium guide, or Transformer Explainer pointing at an in-browser visualization. It is not a tutorial and does not pretend to be one.

How the resource list is generated: resources.ts, the website, and the README

The mechanism is the part worth understanding before you fork it. The README carries an explicit marker in the source: an AUTOGEN block that says do not edit by hand and states the content is generated from website/src/data/resources.ts via npm run gen:readme. So the TypeScript file is the source of truth, not the Markdown. Edit the README directly and the next generation run overwrites you. The repository layout supports this: top-level entries are .github/, .gitignore, CONTRIBUTING.md, LICENSE, README.md and website/. The website directory holds the data file and presumably the site that renders it, which is why the project's primary language is listed as TypeScript even though the visible artifact is a Markdown list. For a contributor this is a good design: one file to change, two outputs updated. For a reader who only wants the links, it means the README is a build product and the live site is the intended reading surface. The homepage is the cheatsheet, and the README itself points there with a Browse the interactive cheatsheet line.

Installing everything-ai-ml and making a first contribution

There is no package to install and no runtime to call. The README does not document an install step for consumers, because there is nothing to consume programmatically. What you clone is the source of the index. The commands below follow from the repository layout and the generation script named in the README's AUTOGEN marker; the README does not spell out a full local setup sequence, so treat the second block as the shape of the workflow rather than a documented recipe.

git clone https://github.com/viveknaskar/everything-ai-ml.git
cd everything-ai-ml

After cloning you get the README, the LICENSE, CONTRIBUTING.md and the website directory. The data you would edit lives at website/src/data/resources.ts. The README names the regeneration command explicitly inside the AUTOGEN marker:

npm run gen:readme

Run that after editing the data file and the README's generated section is rewritten to match. If you only want to read the material, skip all of this and open https://viveknaskar.github.io/everything-ai-ml/, which the README presents as the interactive cheatsheet. The practical first use is a pull request: add one link to the correct section in resources.ts, regenerate, and check the diff touches only the generated block. CONTRIBUTING.md is the file to read before you do, since the README does not restate its rules.

Where everything-ai-ml stops being the right tool

The failure mode is expectation mismatch, and it is a big one. Nothing here executes. If you want to run a RAG pipeline, fine-tune a model, or stand up an MLOps stack, this repository gives you links about those things and no code for them. The TypeScript in the repository builds a website; it is not an SDK. A second limitation is link decay. Every entry is an external URL, and the project has no visible link checker in the top-level layout. The README does not document automated validation of the URLs, so a broken link is a maintenance event someone has to notice and fix through resources.ts. Third, the index is opinionated and unranked. Entries are grouped by topic but the README gives no difficulty ordering, no prerequisite chain and no indication of which links are current versus dated. For a beginner who needs a sequence, the AI/ML Roadmap section is the closest thing to a path, and it is a list of links rather than a syllabus. Finally, the descriptions are one line each. That is enough to decide whether to click, not enough to decide whether a resource is good.

everything-ai-ml against awesome-lists and structured courses

The closest alternative in kind is the awesome-list pattern: a Markdown file of links, edited by hand, rendered only by GitHub. The difference in approach is the build step. An awesome-list is the artifact; everything-ai-ml treats the Markdown as output and keeps structured data in website/src/data/resources.ts, which lets the same entries drive a website with its own presentation. That buys a searchable cheatsheet and costs contributors a generation command. The other alternative is a structured course such as the Harvard CS50 material the README itself links under AI/ML Roadmap, or the MIT OpenCourseWare linear algebra course listed under Building Blocks. Those give sequence, exercises and assessment; everything-ai-ml gives breadth and pointers. A third option is the official documentation of a framework, for example the Scikit-learn pages linked for model evaluation and clustering. Official docs answer how to use one tool. everything-ai-ml answers what exists across the field. Choosing between them is choosing between a syllabus and a map.

Maintenance, licensing and the cost of keeping an index alive

The repository is not archived and the last push was on 2026-09-10, so the index is being touched. The single release listed is v1.0.0 from 2026-04-02. For a resource list, an upgrade is not a version bump you install; it is a pull of main. The real cost sits on whoever maintains a fork. Sections such as Coming Innovations in LLMs and Tools and Frameworks age fastest, and each stale entry is a small editorial decision. The generation step means a fork carries two artifacts to keep in sync, the data file and the README it produces. Licensing is simple on its face: the repository is MIT, so reuse and redistribution of the list are permitted under that licence's terms. One caveat that is not a legal opinion: the links point at third-party content under their own terms, and the MIT licence on this repository does not extend to what those pages contain. Check the destination before you republish anything beyond the link itself.

What the section list tells you about scope

The breadth is the product, and the section names are the honest description of it. A single index covers classical concepts (Supervised Learning, Clustering Algorithms, Bayesian Inference), foundations (mathematics, probability, optimization, feature engineering), generative topics (Prompt Engineering, RAG, Fine-tuning), operational topics (MLOps and GenAIOps, Security, AI Cost Optimization), organizational material (Adopting GenAI in Organizations, AI Augmented SDLC) and career material (Courses, Certifications, Books, Practice Problems, Interview Preparation). That range is unusual for one repository and it is also the weakness: a reader who wants depth in one area gets a handful of links, while a reader who wants orientation gets a lot. The interactive visualizations listed under Key Concepts, such as MLU-Explain, CNN Explainer and Transformer Explainer, are the entries most likely to be useful to someone who has not yet built intuition, because they show the concept rather than describe it. If you are evaluating the index for a team, read the section you care about first and count how many links in it are still live.

Editorial conclusion

Adopt everything-ai-ml if you want a single MIT-licensed index of AI/ML reading to point a study group or a new hire at, and if you are willing to send a pull request when a link rots. Do not adopt it if you need runnable notebooks, versioned course material, or a dependency you can install and call from code; it ships no library and no API. Before relying on it, verify the live site at viveknaskar.github.io/everything-ai-ml still resolves, check that the section you care about is populated in website/src/data/resources.ts, and confirm the last push date on the default branch so you know how fresh the index is. The repository is not archived, and the most recent push was on 2026-09-10.

Frequently asked questions

Official sources

1. License: MIT
2. Project website
3. README
4. Releases
5. viveknaskar/everything-ai-ml on GitHub