everything-ai-ml: A Curated Link Index Built From One TypeScript Data File
A curated collection of learning resources for Generative AI, Machine Learning, Agentic AI, LLMs, RAG, Fine-tuning, MLOps, and more.
At a glance
- What is it?
- The repository is a resource list, not a library. Its README is generated from website/src/data/resources.ts, and the live site is a GitHub Pages cheatsheet. Here is what that architecture means if you are deciding whether to point your team at it.
- Who is it for?
- Adopt everything-ai-ml if you want a single browsable index of AI/ML links and are willing to check each link's current state yourself, since the repository carries no automated link checking that the material reveals. Do not adopt it as a dependency, a course, or a source of vetted technical claims: it is a list, and its value is the editor's selection, not verification.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 1 day ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What Problem a Generated Link Index Solves
The problem is not a missing tutorial. It is that AI/ML learning material is scattered across vendor explainers, university course pages, personal blogs, YouTube channels, and arXiv papers, and a person entering the field has no way to tell which of those are worth an evening. everything-ai-ml answers that with curation: one repository, roughly thirty top-level categories, each holding annotated links. The README shows the shape. Under AI/ML Key Concepts there are entries for supervised learning, reinforcement learning, NLP, computer vision, GANs, dimensionality reduction, clustering, Bayesian inference, time series, and self-supervised learning, each with a one-line description and a source. Some point at vendor documentation (AWS on GANs, Google Cloud on unsupervised learning), some at scikit-learn docs, some at personal posts. The audience is the person who wants a starting map, not a syllabus with exercises. If you already know which papers matter in your subfield, this list will feel shallow. If you are trying to work out what RAG, fine-tuning, and MLOps even refer to before choosing a project, the category structure does that work for you.
The README Is a Build Artifact, Not the Source
This is the detail that separates the repository from the thousands of awesome-list clones. The README contains an explicit marker: AUTOGEN:START, with the comment that it is generated from website/src/data/resources.ts via npm run gen:readme, and an instruction not to edit it by hand. So the canonical content lives in a TypeScript file, the README is derived from it, and the site at viveknaskar.github.io/everything-ai-ml is the interactive view of the same data. That single-source arrangement is the real engineering decision here. It means a contributor adding a resource edits a typed data structure rather than a markdown list, and it means the README and the site cannot drift apart as long as the generation step runs. The cost is that the repository's primary artifact is only as good as the generator. If you fork it and edit the README directly, your change will be overwritten the next time someone runs the script. The material does not show the schema of resources.ts, so how a resource is typed, whether categories are enumerated, and whether link health is stored are all things you would have to read the file to learn. That is the first file to open.
Category Coverage: From Key Concepts to Interview Prep
The table of contents is the most informative part of the README, because it shows what the maintainer thinks the field contains. It runs from foundational material (AI/ML Key Concepts, AI/ML Building Blocks, AI/ML Roadmap) through applied topics (Prompt Engineering, RAG, Fine-tuning, Frameworks, Agentic AI, MLOps and GenAIOps) into operational and organisational ground (Security, AI Cost Optimization, Adopting GenAI in Organizations, AI Augmented SDLC). It also covers vendor-specific material (Google Cloud AI and ML), adjacent fields (Quantum Computing and PQC), and career material (Courses, Certifications, Books, Must-Read Research Papers, Interview Preparation, Practice Problems). Two things stand out. First, the inclusion of cost optimisation and organisational adoption alongside model topics reflects where practitioners actually get stuck, which is usually procurement and rollout rather than architecture. Second, the breadth is a warning as much as a feature: a list covering quantum computing and interview prep is not deep in either. The Building Blocks section is a good example of the intended depth. It lists mathematics for machine learning, linear algebra, probability and statistics, calculus for optimisation, Python, optimisation techniques, preprocessing, evaluation metrics, regularisation, loss functions, activation functions, and hyperparameter tuning, each as one link. That is a reading list, not a curriculum, and the repository does not pretend otherwise.
Running the Site Locally and Regenerating the README
The material gives one command explicitly: npm run gen:readme, described as generating the README from website/src/data/resources.ts. That tells you the toolchain is Node and npm, and that the site lives in a website directory with a src/data folder holding the resource definitions. What the supplied material does not give is the install and dev-server commands, so treat the following as the sequence the layout implies rather than something quoted from the documentation: clone the repository, change into the website directory, run npm install, then run the project's dev script to serve the site locally, and run npm run gen:readme from wherever the package.json defining that script lives. The one thing you can state with confidence is the edit workflow: change resources.ts, run the generator, commit both the data file and the regenerated README. If you are evaluating the project rather than contributing, the practical move is to read resources.ts directly. It is the whole dataset in one file, and reading it takes less time than clicking through thirty categories on the rendered site. The presence of a v1.0.0 release dated 2026-04-02 and a later push on 2026-09-10 indicates the repository is still being updated rather than frozen, though the material does not describe what changed between them.
Where a Static Link List Breaks Down
The failure mode is link rot, and nothing in the supplied material indicates automated checking. A curated list is a snapshot of what was useful on the day each entry was added, and AI/ML moves fast enough that vendor explainers get restructured, course pages move behind enrolment, and blog posts disappear when a domain lapses. The repository shows a v1.0.0 release and a more recent push, which suggests periodic attention, but that is not the same as a CI job that fails when a URL returns 404. The second limitation is depth. A single link per concept cannot tell you whether the linked article is accurate, current, or pitched at your level. The one-line descriptions in the README are selection rationale, not review. The third is that the project is not a dependency and not a course. You cannot import it, and it does not teach anything itself. If you want a structured curriculum with exercises and graded feedback, a list of links is the wrong tool regardless of how well curated it is. If you want a maintained index to hand to a new hire on their first week, it is the right shape.
How This Differs From a Framework or a Course Platform
The obvious comparison is not another awesome-list, which would be the same thing with a different editor. It is a structured learning platform such as fast.ai or DeepLearning.AI, and the difference is architectural. A course platform owns the sequence: lesson one before lesson two, exercises, and an assessment that tells you whether you understood. everything-ai-ml owns none of that. It owns a taxonomy and a set of pointers, and the sequence is whatever you make of it. That is a genuine trade-off rather than a shortcoming. Course platforms are opinionated and slow to update, because revising a lesson costs more than swapping a URL. A data-file-driven index can add a new category the week a topic becomes relevant, which is presumably how Agentic AI and AI Cost Optimization ended up alongside the fundamentals. The other comparison is a documentation site such as the scikit-learn user guide, which the repository links to for clustering and model evaluation. That documentation is authoritative, versioned, and maintained by the library's own team. everything-ai-ml links to it rather than replacing it, which is the correct relationship: the index routes you to the primary source, and the primary source does the teaching.
Licence, Maintenance, and What Updating Actually Costs
The repository is MIT licensed, which is permissive and places few obligations on reuse: you can fork it, modify resources.ts, and publish your own version. Two caveats follow from that, and neither is legal advice. First, MIT covers the repository's own contents, meaning the data file, the generator, and the site code. It does not cover the third-party pages the links point to, and those carry their own terms, paywalls, and attribution requirements. Second, if you fork and republish, keeping the licence file intact is the conventional practice. On maintenance: the cost of keeping a fork current is not the code, which is small, but the review work of deciding which new links belong and which existing ones have gone stale. The generator removes the mechanical part of that (edit the data file, run npm run gen:readme) but not the editorial part. Upstream, the material shows a single maintainer name in the repository path and a sponsorship link in the README, which is worth knowing if you plan to depend on the list staying current. A one-person curation effort with a newsletter attached can move quickly or stall, and the push timestamp is the only signal the material offers.
Who Should Adopt It, and What to Check First
Use it if you are building an onboarding path for engineers moving into AI/ML work and you want a browsable index rather than a course, or if you want a starting point for your own curated list and would rather fork a working generator than write one. The TypeScript data file plus npm run gen:readme pattern is the part worth copying even if you discard the links. Do not use it as a substitute for primary documentation when you are implementing something, and do not treat any entry as an endorsement of the linked page's accuracy. Before you commit to it, do three things. Open website/src/data/resources.ts and confirm the categories you care about have more than a token entry. Check whether the repository has any link-checking step, since the supplied material does not show one, and if it does not, budget for periodic manual review of the links your team uses. Finally, read the README's generated marker and make sure your own edits go into the data file, because editing the README by hand is the one mistake this architecture actively invites.
Editorial conclusion
Adopt everything-ai-ml if you want a single browsable index of AI/ML links and are willing to check each link's current state yourself, since the repository carries no automated link checking that the material reveals. Do not adopt it as a dependency, a course, or a source of vetted technical claims: it is a list, and its value is the editor's selection, not verification. Before relying on it, open website/src/data/resources.ts, confirm the categories you need are populated, run npm run gen:readme after any edit to confirm the README regenerates cleanly, and spot-check the handful of links your team will actually use, because the MIT licence covers the repository's contents and not the third-party pages it points to.
Community notes