Zachary Zhang:

A component library of 1,272 components for GPUI, the Rust UI framework behind the Zed editor, giving Rust desktop applications ready-made parts, each designed for both a light and a dark theme.

It spans primitives, forms, menus, and navigation through to complete application surfaces such as tables, charts, a code editor, a terminal, chat, mail, git views, and finance widgets, enough to build an IDE, a trading desk, or an AI chat application.

Every component runs live in the browser in a 43-chapter gallery at elygpui.com, so developers can try each one before adding the library; open source under MIT.

Responsible for Omarchy's Chinese ecosystem: localization, compatibility with China's network environment, Chinese input methods, related tool adaptation, user feedback, and upstream collaboration.

Earlier, launched OmarchyCN alone, adapting the Chinese environment, package mirrors, the install and update flow, local AI tools, and the related infrastructure for users in China; it reached close to ten thousand system downloads in its first two days, followed by joining the official Omarchy team as the Chinese-region maintainer.

A plugin for the Omarchy 4 bar that plays animated Codex Pets sprite sheets in the bar, in a panel, or pinned and draggable on the desktop, with poses, idle wandering, a pet that follows the cursor, and a wave on click.

Built the QML bar widget, the Python validator and command-line tool for installing and uploading pets, and omarchy-pets.com, a public library where users browse, install, and share pets.

A court-gothic component library for MyGo, the native Go UI: ivory and charcoal grounds, a wine-red accent, old-gold rules, and serif headings, across 100 components in light and dark themes and three densities.

MyGo draws every component on the GPU, with no HTML, JavaScript, or WebView; the gallery runs as a native window or mirrors itself to a local browser. Open source under MIT.

500 pixel-art components in the manner of 16-bit consoles for MyGo native UI, in 22 categories from a button to a chessboard. Every dot sits on the grid, every color comes from a sixteen-slot palette, and nothing is anti-aliased.

Eight palettes, four seasons by day and by night, repaint every component at once; the showcase is the same Go gallery compiled to WebAssembly and drawn dot by dot in the browser.

An AI coding agent built in Rust for multi-agent collaboration, parallel execution, and autonomous development in complex software engineering.

Its core design is Graph Engineering: a high-level engineering goal is automatically broken down into a dependency DAG, and a Planner, parallel Workers in Git worktrees, a Verifier, Replan, and Final Verification carry the work through. Agent Swarm hands large sets of independent tasks to many sub-agents in parallel.

It also has a multi-provider agent runtime that works with subscription accounts and API providers alike, with a TUI, a headless CI mode, ACP editor integration, cross-platform installation, and automatic updates.

Offline-first desktop office apps in Rust on GPUI and Ely GPUI: ElyDoc for documents, ElyTable for spreadsheets, ElyShow for presentations, ElyPDF, ElyTree for Git, and ElyMark for Markdown.

Word, Excel, PowerPoint, and OpenDocument files open and save through LibreOfficeKit; ElyPDF marks up, fills, and signs PDFs on MuPDF and finds text in scans with Tesseract; ElyMark draws math and diagrams in place.

Files stay in the user's own folders, with no account and no cloud service; open source under AGPL 3.0, introduced at office.elygpui.com.

A terminal for macOS, Windows, and Linux in Rust on GPUI and Ely GPUI, with sessions on the left, the terminal in the middle, and Git or Files on the right; the emulator runs on alacritty_terminal with truecolor, the kitty keyboard protocol, and shell integration for zsh, bash, fish, and PowerShell.

Sessions report waiting, errors, progress, and completion through OSC marks and system notifications, and Ely, an animated mascot, peeks over the window edge and reacts to what the sessions are doing; open source under MIT.

A layered, Photoshop-style image editor for Linux, built first for Omarchy and following the active Omarchy theme; a port of Compositor, an open-source image editor for macOS.

Supports layers, groups, blend modes, masks, clipping masks, adjustment layers, and styled text, with 15 tools and 12 adjustment types, and opens PSD, HEIC, camera RAW, and other common formats.

Includes one-click background removal that runs entirely on the device, projects that move freely between OmaPhoto and the Mac application, and one-command installs for Arch, Ubuntu, Fedora, and NixOS; open source under MIT.

A privately deployed ontology runtime for enterprises that maps ERP, MES, WMS, databases, files, and real-time event streams onto unified business objects, relationships, and behaviors.

The system builds the enterprise semantic layer around Object, Property, Link, Function, and Action, so an AI agent can read the state of the enterprise, run analysis, call models, and start business actions, entering real business processes within the bounds of permissions, approvals, and audit.

It also carries an Observed World / Simulated World architecture for simulating and playing out strategy, production, inventory, resource allocation, and other complex decisions.

An AI city operating system for government and urban governance, built on DataOS + AgentOS, that maps population, enterprise, industry, transport, public facility, and city event data from many sources onto unified digital city objects.

Through a city-scale ontology, multi-agent systems, prediction models, and simulation, AI can read the state of the city, analyze problems, play out the results of policies, and generate executable tasks.

The project delivered a proof of concept for the Kuala Lumpur city government in Malaysia and took third place in the CityOS hackathon.

Accountability and audit infrastructure for autonomous AI agents that gives an agent's key operations a verifiable, tamper-evident, traceable record.

Digital signatures, a hash chain, and a responsibility protocol record the agent's identity, action, time, context, and result, so an organization can trace what an AI agent did, why it ran, and what came of it.

The project takes on accountability, auditability, and the attribution of responsibility once agents act inside finance, enterprise systems, and real assets.

An experimental platform for autonomous multi-agent organizations that defines Company, Department, Role, Agent, Task, Decision, Memory, Policy, and Approval in software.

Many AI agents collaborate over the long run by organizational structure, job responsibilities, and permissions, completing company-level goals through task assignment, execution, review, and replanning, while a policy engine and human approval control high-risk operations.

The project explores Organization-as-Software, long-running multi-agent collaboration, permission boundaries, organizational memory, and the autonomous company.

An identity and authorization system for AI agents that settles who an agent is, what it may do, and whether its execution can be trusted when it calls APIs, tools, enterprise systems, and other agents.

Its core directions are agent identity, authentication, authorization, credentials, policy, and verifiable machine identity, a base authentication layer for the large-scale agent-to-agent and agent-to-service interaction to come.