Matthew Berman demonstrates several generated browser applicationsCode generation uses AI or another automated system to create source code from instructions, examples, schemas, or higher-level specifications., including an interactive mechanical watch, a game and a foldable phone model. He notes imperfections in the game and animation while giving his assessment of their visual quality.
Berman tests the model on counting and estimating partially hidden pastries, selecting interview clips and producing a spreadsheet-like application locally. The visual countingA multimodal model can process or generate more than one kind of data, such as text, images and audio. estimates and his assessment are presented as examples from his tests, rather than independently verified performance measurements.
He also demonstrates a PDF reader generated from a detailed prompt and discusses benchmark resultsA benchmark is a defined test or collection of tasks used to compare a system's performance under specified conditions., inference speedAI inference speed measures how quickly a trained model begins and continues producing outputs for a request. and pricing. His overall ranking is a personal judgment, and the examples do not establish performance across every task or deployment.
Watch on YouTube
