# One YAML file, five open to-do items, and an inclusion criterion the author is still thinking about > stared/interactive-machine-learning-list is a hand-curated directory of interactive machine learning, deep learning, and statistics websites, rendered as a no-build Vue.js site from a single websites.yaml. What it does well is say who may add to it; what it has not done yet is describe, sort, or release anything. **stared/interactive-machine-learning-list** — A collaborative list of interactive Machine Learning, Deep Learning and Statistics websites - Repository: https://github.com/stared/interactive-machine-learning-list - Website: https://p.migdal.pl/interactive-machine-learning-list/ - Stars: 455 · Forks: 43 - Language: JavaScript - License: MIT - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/stared-interactive-machine-learning-list ## The admission criterion is still an open question in the project's own text The section that decides what belongs opens by admitting it has not decided: Still I am thinking what is the best criterion. One rule is firm, front-end work, meaning JavaScript running within the browser. Everything that needs a server is explicitly unsettled, and the stated lean is toward inclusiveness, which is a generous default for a directory. Two conditions come attached. An entry has to have didactic value, with the parenthetical that otherwise ALL services using ML would qualify, and it has to add `backend-dependent` in `uses`. That field is the only machine-readable part of the policy: a tag in the data file that a reader, or a future filter, can act on. Everything else about the boundary lives in prose in the README, which means inclusion is decided by whoever reviews the pull request rather than by a script. ## One YAML file is the entire submission path Contributions arrive as pull requests against a single file, and the README points contributors straight at `websites.yaml` on the master branch. There is no submission form, no command-line tool, and no schema file in the layout. The top level of the repository is short: `.github/`, `.gitignore`, `LICENSE`, `README.md`, `imgs/`, `index.html`, `screenshot.png`, `src/`, and `websites.yaml`. Notably absent are `package.json`, a lockfile, and a test directory, which fits the description of the site as a simple no-build Vue.js website with nothing to install before opening it. The consequence for a contributor is that the YAML is the whole contract, and the layout does not spell out how that file reaches the browser at runtime. ## The to-do list says entries have no descriptions and no sorting yet The project publishes its own backlog, headed by an invitation that is misspelled constribute, and it is worth reading as a feature list with several items unchecked. The five entries are descriptions of sites, a write-up in a different way, some sorting, with alphabetical offered as a question mark, a share button, and code refactoring. Two of those are about what a visitor gets: there are no descriptions attached to the entries, and the order is whatever the page currently uses rather than anything a reader can predict. That also sets the ceiling on the curated claim. Without descriptions or sorting, the list tells you that a site exists and where to find it, and leaves the judging to you. ## The canonical copy sits on a personal domain, not on a GitHub page The address the project gives as its homepage is `https://p.migdal.pl/interactive-machine-learning-list/`, a path on the maintainer's own domain, and the README links the same URL as the live site. The default branch is `master` rather than `main`, which matters in practice because the YAML link in the README is written against that branch name. There are no GitHub releases, so the repository has no versioned artifacts of any kind, and the repository itself is the source of truth rather than a packaged site. For anyone who wants to build something similar, the project points elsewhere rather than shipping a template: the README sends you to In Browser AI if you want to create such visualizations yourself. ## Four sibling lists and two credited sources place it in a lineage The README situates the directory by linking outward twice. Under other lists it points at Explorable Explanations, Distill, Explained Visually, and AI Experiments with Google, four established venues in the same space. Under inspirations it names Explorable Explanations by Bret Victor, describes a collaborative science-based-games list as one the author started, and adds a parenthetical about maybe turning it into something interactive as well. Two further entries, a searchable compilation of Kaggle past solutions and D3 Discovery, each carry a source link to the underlying repository. So the collection side is explicit about where it learned its taste, while the submission side is the half that runs on prose. ## Open source is a preference, and licenses are named only when they matter The stated policy is a strong preference for open-source solutions, so that people can reuse the work and learn from the code, and then an explicit statement that it is not a requirement. A closed service can be listed. The rule that follows is more interesting: mention the repo and the open source license only when it is directly relevant, with the contrast drawn against additional materials such as exercises for a book or a Python algorithm. In practice that means a license line appears on some entries and not others, on purpose. The repository carries its own MIT license in a root `LICENSE` file, which covers this directory; the linked pages keep their own terms, and the list is careful not to speak for them. ## The most recent push is dated 2026-03-15, with no releases behind it The repository is not archived, and it carries an MIT license, a JavaScript primary language, 455 stars, 43 forks, and 3 open issues. What it does not carry is a release, a tag, or a changelog. For a static site whose entire content is one YAML file, the commit date is the only upkeep signal a reader gets, and the latest one on the default branch is dated 2026-03-15. That is a fact worth carrying into any use of the list: there is no way to pin a version, no way to see what changed between two visits, and no artifact to diff against when a linked site starts behaving differently. ## Conclusion Use this list if you want a curated starting point for browser-based ML explanations, and expect to hit the same wall the to-do list names: no descriptions, no sorting, and no entry history beyond commit dates. Before you rely on it, check the three things the project leaves open. Its inclusion criterion is described as still undecided, its ordering is not alphabetical, and its most recent push is dated 2026-03-15, so a page you liked may have changed since you saw it linked. If you want to add a site, the pull request path is short and the rules are in the README, but a backend-dependent entry needs the `backend-dependent` tag in `uses` and a didactic justification. ## FAQ ### What goes into stared/interactive-machine-learning-list? Front-end work, meaning JavaScript running within the browser. Entries that need a server can still qualify, but the project asks that they carry `backend-dependent` in `uses` and have didactic value, since otherwise all services using ML would qualify. ### How do I add a site to stared/interactive-machine-learning-list? Open a pull request against `websites.yaml` on the master branch. The top-level layout holds no package.json, no submission form, and no schema file, matching the description of the project as a simple no-build Vue.js website. ### Does every entry in stared/interactive-machine-learning-list have to be open source? No. The project states a strong preference for open-source solutions so people can reuse them and learn from the code, and then says outright that it is not a requirement. It also asks that a repo and its license be mentioned only when directly relevant. ### Does stared/interactive-machine-learning-list describe the sites it lists? Not yet. Descriptions of sites, a write-up in a different way, some sorting, a share button, and code refactoring all sit in the project's own to-do list, so an entry is currently a name and a link. ### Where is the live version of stared/interactive-machine-learning-list? At https://p.migdal.pl/interactive-machine-learning-list/, the address the project gives as its homepage. The repository has no GitHub releases, and its default branch is `master` rather than `main`. ### How current is stared/interactive-machine-learning-list? The default branch was last pushed on 2026-03-15 and the repository is not archived. It carries an MIT license, 455 stars, 43 forks, and 3 open issues, with no releases, tags, or changelog to compare against. ## Sources - [Issues](https://github.com/stared/interactive-machine-learning-list/issues) - [License: MIT](https://github.com/stared/interactive-machine-learning-list/blob/master/LICENSE) - [Project website](https://p.migdal.pl/interactive-machine-learning-list/) - [README](https://github.com/stared/interactive-machine-learning-list/blob/master/README.md) - [stared/interactive-machine-learning-list on GitHub](https://github.com/stared/interactive-machine-learning-list) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/stared-interactive-machine-learning-list