AI products are taking the form of a host app, a licensed catalogue and a single file
Google is consolidating features in Gemini, Pippa is building a licensed style catalogue, and community demonstrations are distributing working game prototypes as editable HTML files.
Artificial Intelligence··Midday
Google is moving from a separate app into its main app
Google cancelled the standalone AI Studio mobile app announced in May 2026 before it reached users. According to the company, hundreds of thousands of pre-orders had accumulated across the Android and iOS stores. The AI Studio account said the capabilities would be combined with the Gemini team across mobile and desktop rather than asking users to install another app. Developer-oriented functions are therefore moving from a separate icon and download channel into Google’s existing artificial intelligence product. The announcement does not say which AI Studio capability will arrive first, which countries will receive it or which subscription tier will include it. Nor does it say that the pre-orders will automatically become Gemini users. The statement also does not explain how user data would move through the transition. The development changes the delivery container: the standalone app will not ship, while its capabilities are intended to continue inside a larger host product.[1]
Pippa is organising the product as a licensed artist catalogue
Pippa is choosing a different form of distribution. The text-to-video service pays an illustrator whenever a subscriber generates an image or clip derived from that artist’s style. The company has licensing agreements with four illustrators and about 800 paying subscribers. Disclosed payments are $0.005 per image and $0.003 per second of video, plus a share of a royalty pool funded with 5 percent of subscription revenue. Monthly plans are offered at several price levels. Illustrators pass a multi-stage process intended to verify ownership of submitted work, may participate under a pseudonym and receive a profile page directing users to their work elsewhere. Pippa says it ultimately wants to rely only on models trained with datasets supplied by participating artists, while its current technology still rests on open models pretrained on a broader body of content. The duration of the licences and the pace at which the catalogue may expand were not disclosed. The product container is a subscription catalogue in which licensed styles become selectable offerings.[2]
Game demonstrations carry the whole product in one file
In community demonstrations using Claude Opus 5, the distributed object is an editable HTML file that runs in a browser. Users reported generating a first-person shooter, submarine game, kart racer and Minecraft-like scene from one prompt apiece. Geometry is created procedurally by code at runtime; textures are shaders, while physics and controls are parts of the same generated code. Running the package on Three.js makes it possible to place the experience in one file instead of loading external assets separately. The creator of one Minecraft demonstration said the build contained 15 biomes and consumed 25 million tokens. Posts also compared outputs from several models, although none held the prompt, number of runs and scoring criteria fixed together. The demonstrations are therefore examples of a distributable format, not a shared performance test. Across the three developments, the place where a capability reaches the user has become part of the product decision: an established host app, a licensed subscription catalogue or a file that can be opened and edited directly.[3], [1], [2]
