The useful corner of Google that most people never find: research previews, free notebooks, teaching tools and public data sets. Almost everything here is free with a Google account.
The front door to everything Google is testing in public, including the image, music and video experiments. Things appear and disappear here without warning, so do not build a workflow on one.
The best thing to come out of Labs so far. Grounds every answer in documents you upload, cites them, and turns the pile into an audio summary you can listen to on a walk.
Google general assistant, strongest when it can reach your Gmail, Docs and Calendar. The free tier is more capable than most people assume.
Prototype prompts against Gemini models, then export the code. The free quota is unusually generous, which makes it the cheapest place to learn how these models behave.
Free Python notebooks with a GPU attached and nothing to install. Sessions time out and hardware is not guaranteed, which is the trade for costing nothing.
The cloud development environment that grew out of Project IDX, with AI assistance and Firebase deployment built in. Handy for full-stack prototypes from a browser.
An asynchronous coding agent that clones your repository, works on a task and opens a pull request. Best for small well-defined chores you do not want to context switch into.
Train an image, sound or pose classifier in a browser in about five minutes, with no code. The single best way to show a room of beginners what machine learning actually is.
Indexes public data sets from governments, universities and repositories in one search box. Underused, and often faster than hunting through a statistics agency website.
Relative search interest over time and by region, which is a decent proxy for demand when you have nothing else. Remember the numbers are normalised, not absolute.
The widest net for academic literature, including theses and grey literature the specialist databases skip. Coverage is broad and the metadata is messier than Semantic Scholar.
Charts how often a phrase appeared in books over two centuries. A niche toy that becomes a real research instrument once you understand the corpus limits.
Gigapixel scans of museum collections plus a pile of interactive experiments. Genuinely useful for teachers and anyone sourcing public domain reference images.
Satellite and historical imagery in the browser, with measurement and project tools. Journalists and investigators use the historical layer more than the 3D view.
An archive of small web experiments with the source code attached. Treat it as a browsing gallery for ideas rather than a set of maintained products.