SYNAPSE
SYNAPSE is an interactive neural-network laboratory. Tap or drag to scatter cyan and magenta points, then hit Learn to watch a tiny neural net discover the curved decision boundary live — complete with an animated loss curve, accuracy readout, and a preloaded interleaved-spiral dataset that forces a non-linear fit.
More neural-network creations

Boundary Lab
Train a tiny neural classifier and inspect its computed probability as a deforming Three.js decision landscape.

Aegis Chorus — Self-Advancing Overseer
An autonomous in-app overseer that plans its own goals, spawns worker agents, and renders internal chatter, agent interactions, and external signals as a live particle network on canvas.

J.A.R.V.I.S. — AI Command Center
A futuristic cyberpunk AI assistant with voice & text commands, a live visualizer ring that reacts to listening/processing/speaking, an app launcher protocol (YouTube, Instagram, TikTok + custom apps), system telemetry (battery, network, clock) and quick tools — notes, clipboard vault and web search. All local-first, no backend required.

Aegis Chorus — Self-Advancing Overseer
An autonomous in-app overseer that plans its own goals, spawns worker agents, and renders internal chatter, agent interactions, and external signals as a live particle network on canvas.

Verbatim Studio — AI Script-to-Movie Maker
Paste any screenplay and Verbatim Studio turns it into a living movie. It keeps every single word — never summarizing or shortening — reads each line aloud with device voice AI, animates the cast to follow the script's actions (enter, walk, run, sit, turn, emotes), autosaves your project as you type, and exports lossless video, .srt subtitles, the project file, and shares straight to your phone's Files/Drive.

PIP — your little talking robot
PIP is a pocket robot pal that genuinely talks out loud. Type a message or tap the mic and PIP chats back, tells jokes, does quick math, reads the time and keeps you company — all with a friendly animated face and a robot voice powered by your browser's built-in speech engine. Works in portrait or landscape, no backend, fully client-side.