Paste a Wikipedia link or search by name, and Erask takes the date and the place from Wikidata. What it cannot find, you fill in — and you never type a coordinate: you drop the point on the globe where the thing happened.
Slower, and exact. Right when you know the subject and care which twenty articles tell it.
One prompt into ChatGPT, Claude, Gemini or whatever you already use. Name a subject and it answers with a link: up to fifty events, resolved against Wikidata so the dates and places are real rather than remembered.
A minute, and approximate. Right when you want to see something you have not thought about before.
Any story can be edited — including one you did not make
This is the part people miss. A story lives entirely inside its own link: there is no copy on a server, no owner, no permission. So a link somebody sent you is not a finished thing you can only watch. Open it, press Edit, and every event is in front of you.
Drop the ones that do not belong. Fix a date. Move a point the model put in the wrong country. Add the three articles it missed. Copy the new link — and that link is yours, while theirs still works, untouched, because you never had it in the first place. You had a copy all along.
It works the same on a story a model wrote and on one somebody built by hand, because after they are written they are the same thing: the same grammar, the same link, the same globe. The only stories that cannot be edited are the ones in our catalogue — those are checked by hand and live in files, so opening one here would make a disconnected copy rather than a correction.
That is also why the honest order is usually: let a model draft it in a minute, then spend ten minutes taking out what does not belong. Starting from something is much easier than starting from nothing.
What the three have in common
No account, no upload, no server. The story travels in the link. Lose the link and the story is gone, so copy it early.
Only identifiers are. The text arrives from Wikipedia when the story is opened — in the reader's own language, and current rather than frozen.
The captions are the expensive part: they travel as text and do not pack down the way identifiers do. Ten of them leave room for hundreds of events; one on every event stops you at under two hundred. Past three hundred the timeline turns into a comb anyway.
Where to go from here
How adding an article works, why most need nothing from you, and how you put a point on the globe without knowing what a coordinate is.
The prompt itself, which models can run it, what it can and cannot do, and how to read the note it writes you at the end.
Two other ways of finding things worth putting in a story: Timed sync, everything on the globe that intersects one year, and the rabbit hole, every place an article mentions.