fak10
Anna's Pensieve
Anna’s Pensieve is a view of the facts and connections she knows, shown as a graph. In this video, you can open Bremen, follow underlined relations, return to a previous node or search for Hamburg. The eleven statements and family are fictional; a missing connection means only that this small example does not contain it.
An inspiration for fak10 was Dumbledore's Pensieve: memories rest side by side, and when you dive in, you see how it all connects. For Anna that's her knowledge store, her Pensieve, your ontology. In this video Anna herself explains how to wander through it.
How do you explore Anna's knowledge graph?
Tap a node to see related facts, follow underlined names to move through the graph, use the back arrow to retrace your steps, or search for a place by name.
- Every dot is something Anna knows. Its color tells you what it is: yellow are people, green are places, light blue are concepts like “city”, dark blue are single things and events.
- Tap Bremen and Anna flies there. Next to it is what she knows about Bremen, and the green box shows what points to Bremen.
- Every underlined name leads on, because every predicate is a path. The small back arrow takes you back, step by step.
- If you know where you want to go, search: type “Ham” and Anna flies to Hamburg.
The world in the video is made up: eleven sentences about an invented family and their cities.
What did filming uncover about the graph?
Filming exposed four interface bugs: a clipped node type, missing names on places and concepts, inconsistent colors for individual things, and repeated names in the node panel. The app was changed to fix each one.
As in episode 2, the video showed bugs in the app that nobody would have noticed otherwise:
- In the node panel, the line with the kind of node, like “place · city”, was cut off at the top.
- In the overview only people carried their names. The video ended in Hamburg, but the only readable name was “Merle”. Now places (green) and concepts (light blue) carry their names too.
- The color for single things depended on the order the data arrived in. Now it is fixed.
- When the panel's own node is the subject of a line, a dash now stands in for repeating its name.
Anna's voice is AI-generated (ElevenLabs). Picture and music are real: a recording of the app in the simulator, and the music is mine.
If you want to be there when fak10 launches: Join the waitlist
How do you follow a path through the graph?
A dot represents a person, place, concept or individual thing. Its color helps tell you which kind it is; its links come from statements. Tap Bremen, for example. The node panel shows what Anna knows about Bremen, and the green box shows statements that point to it. Tap an underlined name or predicate to follow the next part of the connection. The small arrow takes you back to the previous view.
This matters when an answer takes more than one step. Instead of reading “Bremen” as a bare result, you can follow the path through the statements and check whether it comes from what you entered. If you already know the place you want, you do not have to start at the beginning of a path: searching for “Ham” takes you straight to Hamburg. Going back and searching again are two ways of moving through the same set of knowledge.
Is Anna's graph a complete map of the world?
No. The cities and family in the video are made up; the test world consists of eleven sentences. The graph shows only the things and connections entered from those sentences or inferred by the app's rules. If a connection is not visible, that does not prove it does not exist in real life. It only means it is not part of this example's knowledge.
The colors are not ratings of importance or privacy either. They mark the kind of node: a person, place, concept or individual thing. The named predicate carries the relationship. In the video you can see what a dot represents, which statements point to it, and how to move through them. The demo shows navigation and display; it is not a complete source of information about real people or places.
What do you find when you open a node?
A node panel takes you from the whole graph to the statements about one person or place. From there, follow an underlined connection or use the arrow to return to the previous view. The video demonstrates this with Bremen: open the place, inspect the statements about it, then move on through the links. Search helps when you know a name but do not know where it sits in the graph.
The view is still only a window onto recorded knowledge. A green place node does not mean Anna has a complete account of that city; it means she treats that node as a place. A missing connection is not a disproof either. To check a conclusion, look at which statements were entered into the example and which could be inferred from them.
Open a node panel and you can also inspect individual facts, including their source and date. A fact's visibility is marked separately. That means the graph does not only lead to another name; it takes you back to the entry supporting a connection. The people and places in the video are fictional, so there are no real sources about Bremen or Hamburg in this test world. The recording uses a small example to show how navigation works.
For background on knowledge graphs, the W3C RDF Primer explains nodes and named links; the SPARQL specification shows how graph patterns can be queried. Anna’s view is not presented as an RDF or SPARQL interface.