lianghsun
Liang-Hsun Huang is an independent programmer and AI scientist focused on applying AI to scientific and legal domains, including drug discovery, NLP, and model fine-tuning. His open-sheet project applies that agent-native approach to spreadsheets: an agent writes React source while the framework handles cell addressing, formula references, recalculation, and export.
At a Glance
- Coding-agent and AI-agent developers
- Engineering teams generating spreadsheets from source/data pipelines
- Finance and operations teams producing repeatable models and reports
- Users of Excel, Google Sheets, and LibreOffice who need live formulas rather than static exports
- +1 more
AI Tools by lianghsun
(1)
open-sheet
AI Spreadsheet Framework for Agents
Discussions
No discussions yet
Be the first to start a discussion about lianghsun
Latest News
open-sheet.dev landing page launched and deployed to Cloudflare Pages
Framework site separated from the open-source repository and repository/package links updated
Version 0.2.0 package release published for core, CLI, and MCP packages
Formula coverage and recipient affordances expanded to 127 whitelisted functions, spill, validation, protection, notes, printing, charts, sparklines, and appendable tables
Products & Services
Open-source spreadsheet framework built for coding agents. Users describe a workbook as React/TypeScript source; open-sheet performs coordinate-free cell references, automatic placement, formula AST serialization/evaluation, live .xlsx generation, and CSV/HTML/PDF export. It includes a viewer, inspect mode, skills/scaffolder, and an MCP server.
The npm core package containing the compiler, placement and reference system, formula engine, viewer, Vite plugin, and open-sheet runtime/CLI functionality; version 0.2.0 is publicly published.
The npm scaffolding/command-line package; its documented entry point is npx @open-sheet/cli init my-sheets.
The npm MCP server package that exposes open-sheet operations over Streamable HTTP so agent frameworks can drive the development server and inspect/describe workbooks.
Market Position
open-sheet positions itself as a source-first, agent-native spreadsheet compiler rather than an assistant that edits an existing workbook. Its stated contrast is Claude for Excel: Claude for Excel helps a person work inside an existing file, while open-sheet is intended to generate the spreadsheet output of a repeatable pipeline and eliminate fragile hand-authored A1 addresses; open-slide and open-doc are adjacent medium-specific frameworks rather than direct spreadsheet competitors.
Leadership
Founders
Liang-Hsun Huang
Independent programmer and AI scientist. His public profiles describe experience spanning web technologies, cybersecurity, networking, deep learning, TensorFlow, PHP/Laravel, penetration testing, and drug-discovery work using TensorFlow and RDKit; he is also identified as the founder of Twinkle AI and founder/community lead of Taiwan Llama.
Founding Story
open-sheet was started around the observation that coding agents are good at analysis but unreliable at hand-authoring spreadsheet coordinates such as =SUM(B2:B13). Its initial vision is to let an agent describe a model in React/source code while the framework owns layout, references, recalculation, and live workbook export, so regenerated spreadsheets remain correct when data grows.
Business Model
Revenue Model
The framework is distributed as MIT-licensed open-source software and its npm packages are public; the inspected sources show no paid plan, subscription, API charge, or commercial revenue model.
Pricing Tiers
The repository and framework are available under the MIT License; the documented setup uses Node/npm or pnpm.
Target Markets
- Coding-agent and AI-agent developers
- Engineering teams generating spreadsheets from source/data pipelines
- Finance and operations teams producing repeatable models and reports
- Users of Excel, Google Sheets, and LibreOffice who need live formulas rather than static exports
- Traditional-Chinese NLP and AI research communities for Huang's separate model projects
- Generated monthly board packs and financial models
- Regenerating workbooks when data lands
- Batch production of invoices and quotations
- Departmental budget-versus-actual reporting
- SaaS KPI dashboards and cohort/conversion analysis
- Fixed-asset registers and depreciation schedules