LumenHarbor

Most software wants your attention. Ours wants to give something back and then get out of the way. LumenHarbor Studio makes small educational applications for Android — the kind you open because there is something you do not yet understand, and close again once you do.

The lighthouse is not decoration. It is the whole brief: hold a light steady on one spot until whoever is looking can finally see the rocks.

We work on subjects that are usually taught badly. Not because the teachers are poor, but because the medium is wrong — a static diagram of a sorting algorithm is a photograph of a river. The thing worth understanding is the motion, and motion needs a screen that moves.

The device does the work. Every application we publish runs its computation on the phone in your hand. There is no server of ours in the loop, which means no outage, no breach of a database we should never have built, and no version of your progress sitting somewhere you cannot reach.

No account, ever, for anything that does not need one. A registration wall in front of a teaching tool is a toll booth on a footpath. If a feature genuinely requires identity it can wait until it does, and so far none has.

Show the real thing. Teaching software that simplifies its subject into something untrue teaches the untruth. When we display code, it is code that runs. When we display a data structure, it is the structure the code is actually operating on, changing in the same instant.

Every permission is borrowed, not taken. The manifest is read line by line before each release. Anything we cannot defend in a single plain sentence gets removed, along with whatever feature asked for it.

Work in progress, without dates attached, because dates on a two-person studio are decoration. Graph traversal — breadth-first and depth-first walks with the frontier drawn as it grows — is written and being fitted into the existing player. A pathfinding module covering Dijkstra and A* exists as notes and one very ugly prototype. Beyond that we are looking at binary search trees, and beyond that we are not looking yet.

Roughly one release every few months. A quiet month means nothing broke and nothing was missing.

Occasionally something gets abandoned. A recursion visualiser took three weeks and never became clearer than a whiteboard, so it was deleted rather than shipped. That is a normal outcome and not a failure.