PLAYGROUND

A controlled comparison of vanilla RNN and LSTM models across two literary corpora. The LSTM converged lower on both datasets and preserved more convincing punctuation, rhythm, and local structure.

1.703

Best test loss

2.458

Best BPC

4

Controlled runs

A set of mobile route states connecting the Figma interface study to working iPhone route previews on the UC San Diego campus.

Low-cost interface studies used to test hierarchy, sequence, and screen density before committing to visual finish.

A mobile interaction study that moves from component behavior to a navigable, testable prototype.

A map of the locally verified TritonNav stack, from the mobile interface and campus data layer to the Valhalla routing service.

One of the real iPhone route states used to examine destination resolution, route legibility, and ETA behavior in the field.