MUZHEN

Projects

Products, tools, and experiments with a clean surface area.

A curated selection of what I’ve built so far — the strongest work first, then smaller archived utilities that still deserve a good home.

7 projects3 featured firstindividual intros included

01

Selected

Meowbert AI

A task-first AI workspace for chats, tools, files, search, and more operational workflows.

Designed around persistent tasks instead of disposable prompts.

Combines chat, tools, files, search, and notifications in one product surface.

02

Selected

Brawl Stars Drafting Engine

A competitive drafting assistant for Brawl Stars with model-backed picks, bans, and win probabilities.

Recommendations are backed by training on millions of top-level games.

Supports searchable brawler selection across picks and bans.

03

Selected

TimeScroll

A privacy-first macOS app for searching your screen history with timeline browsing, OCR text search, and local AI recall.

Timeline view makes it easy to jump back to any moment visually.

Searches text that appeared on screen and highlights matches directly.

04

Archived

folder2txt

A browser-based utility for turning folders, ZIPs, and repos into clean text output without uploading data elsewhere.

Accepts folders, ZIP files, and Git repositories as input.

Lets you expand, collapse, and shape exported structure before download.

05

Archived

raycast-g4f

A Raycast extension that brings free or custom AI models, chat, web search, uploads, and image generation into Raycast.

Supports multiple providers and a broad range of language models.

Includes chat, web search, file upload, and image-generation workflows.

06

Archived

File Drive

A lightweight personal file storage service with accounts, uploads, and hosted file management behind a minimal web UI.

Account-based access with login and sign-up flows.

Built around straightforward upload, storage, and file management tasks.

07

Archived

Wordle Solver

A small solver utility for narrowing down Wordle guesses faster with lightweight feedback-driven filtering.

Built around quick iteration instead of heavy setup.

Uses feedback from prior guesses to narrow candidate words.

Open introductionLink coming soon