Public-domain books, proofread by people rather than scanned and dumped. The formatting is plain and the catalogue is the classics, which is the point.
Resource sites that stay useful
Places to get files, references and archives without a subscription or a legal question mark. Ranked on how long they have been reliable rather than on how much they hold.
Takes the same public-domain texts and typesets them properly, with modern covers and clean metadata. The difference over the raw scans is obvious on the first page.
The media library behind Wikipedia, and the first place to look for an image whose licence you can actually read. Every file states its terms and its source.
Unsplash
UnsplashPhotographs, free to use including commercially, with no attribution required. The catch is that everyone else has the same photos, so it is a starting point rather than a brand.
Pexels
PexelsThe same idea with a larger video section, which is the reason to check it alongside Unsplash rather than instead of it.
Pixabay
PixabayImages, video, music and sound effects in one place under the same licence. Quality varies more than the dedicated sites, and the convenience is real.
OpenStreetMap
OpenStreetMapMap data that is open rather than licensed per use, which is why it turns up inside everything from delivery apps to humanitarian logistics. Editing it is allowed, and often the point.
Our World in Data
OxfordLong-run statistics with the underlying data downloadable and the methodology written out. The charts are the shop window; the CSV files are the product.
Kaggle
GoogleDatasets plus notebooks that show how other people handled them. The notebooks are the more underrated half: a working example beats a data dictionary.
Hugging Face Datasets
Hugging FaceThe default place to get a dataset into a script, because the loading is one line. Read the licence on each one; they differ more than the interface suggests.
arXiv
CornellPreprints before peer review, which is both its value and its hazard. The rate of publication is the point: in fast-moving fields the journal is already late.
Full text for the biomedical papers that are open, which is a large and growing share. The place to go when the abstract is not enough to answer the question.
Zotero
ZoteroReference manager that captures a citation from the browser tab you are reading. Open source and it works offline, which matters more than the feature list.
Calibre
CalibreThe ebook library manager that also converts between every format worth naming. Ugly, complete, and the only tool that does the whole job.
AlternativeTo
AlternativeToCrowdsourced replacement lists for software you are trying to leave. The signal is noisy and the coverage is the widest there is.
Product Hunt
Product HuntWhat launched this week, which is exactly as useful as that sounds: good for awareness, unreliable as a quality signal because the ranking is a launch-day game.
Music cleared for reuse, with the licence on every track. The catalogue is not current pop and it is not trying to be.
Freesound
Universitat Pompeu FabraSound effects uploaded by the people who recorded them, each with a stated licence. Searchable by how a sound is described rather than by name, which is the hard part.
Simple Icons
Simple IconsBrand marks as plain SVG files, free to use. The reason a directory site can show a logo beside every entry without touching anyone’s asset kit.
Lucide
LucideA consistent interface icon set with one stroke weight across the whole library, which is the thing hand-picked icons never manage.
Sources and method
The ranking is ours. Anything on this page that can be checked is checked against a named source, and the date we last opened that source is printed with it. If a figure and our verdict disagree, trust the figure and follow the link.
No site is listed twice. A listing means we used it for a month first, and a free tier only counts as free if it survives the trial period.