Slowly
Online

Catch up on the group chat without reading it.

Paste a long conversation. A language model running inside this page gives you the gist, what you need to do, and the dates that matter. Your messages never leave your phone.

Model
Not loaded
01

Chat

0 charactersStays on this device
02

Summary language

Load a model first. Tap Set up above.

Saved

Your summaries, kept on this phone only.

Lab

Choose an open-weight model, load it, and measure how it runs on this phone.

01

Model

02

Benchmark

Runs a fixed Hinglish test chat on the loaded model and records load time and speed. Switch models and run it again to compare.

03

About

How it works

Your browser downloads a compressed open-weight model once and stores it. After that, the model runs on your phone's GPU through WebGPU. Long chats are read in parts and combined.

Why open weights

The weights are public, so the model runs here with no account, no API key and no per-use fee. Your chat is never sent to a server, and you can switch models.

Limits

Small models can miss details or misread tone. Treat the summary as a quick catch-up and check the original chat before acting on anything important. Needs a recent Chrome on Android with WebGPU.

Built with open-weight models for the Hacktoberfest 2026 Open-Source AI Challenge.