@ShenHuang: Testing the cultural confidence of the Chinese-speaking circle. An author of a DeepSeek TUI project posts a couple of Chinese sentences on X, visits China, and the project gets 30k stars. I have studied Chinese for decades and return to China countless times each year. I work on Harness in Google's AI mode. The DeepSeek TUI distilled from experience...

X AI KOLs Timeline Tools

Summary

The author released an open-source TUI coding agent tool called OpenSeek, which supports multi-provider routing and rich features, while expressing dissatisfaction with the cultural confidence of the Chinese-speaking circle.

Testing the cultural confidence of the Chinese-speaking circle. An author of a DeepSeek TUI project posts a couple of Chinese sentences on X, visits China, and the project gets 30k stars. I have studied Chinese for decades and return to China countless times each year. I work on Harness in Google's AI mode. The DeekSeek TUI - openseek distilled from experience only has 5 stars, and one of them is from myself and a friend. Why
Original Article
View Cached Full Text

Cached at: 05/17/26, 01:25 AM

Testing the cultural confidence of the Chinese-speaking world

The author of a DeepSeek TUI project posted a few lines of Chinese on X, visited China. The project then got 30k stars.

I have been learning Chinese for decades, and I return to China countless times a year.

I work on Harness in Google’s AI mode. After distilling the experience, I made DeekSeek TUI - openseek, which only has 5 stars, and one of them is from myself and my friend.

Why


LichAmnesia/openseek

Source: https://github.com/LichAmnesia/openseek

OpenSeek - Broad Exploration

Chinese: README-zh.md Website: https://openseek.dev/

Open-source TUI coding agent — a TypeScript / Bun monorepo.

Lightweight terminal coding companion with multi-provider routing (OpenAI-compatible / Anthropic / Bedrock / Vertex / Azure), 50+ tools, 100+ slash commands, Plan / Agent / YOLO modes, MCP client, LSP feedback, and a headless HTTP/SSE runtime API.

Requirements

  • macOS or Linux (Windows: use WSL2)
  • Bun (https://bun.sh) ≥ 1.3 — required at runtime, not optional. The codebase uses Bun.spawn / Bun.serve / Bun.file; node alone cannot run it. Even if you install via npm i -g, the openseek command shells out to bun at startup and will fail with env: bun: No such file or directory if bun is not on your PATH.

Install

Step 1 — install bun (skip if already installed)

curl -fsSL https://bun.sh/install | bash   # any Unix
brew install oven-sh/bun/bun               # macOS (Homebrew)

Verify: bun --version should print ≥ 1.3.0.

Step 2 — install openseek

Pick one. All three put openseek on your PATH.

# A. npm
npm install -g openseek

# B. bun (same registry as A, smoother if you already use bun)
bun add -g openseek

# C. From source (recommended while iterating)
git clone https://github.com/LichAmnesia/openseek.git
cd openseek
bun install
bun run build
ln -sf "$PWD/bin/openseek" ~/.local/bin/openseek

Step 3 — run

openseek               # start the TUI
openseek doctor        # health check
openseek serve --http  # headless HTTP/SSE on :7117

Develop

bun install
bun run dev          # boots the CLI directly from TypeScript sources, no build needed

Verify

bun run lint        # biome
bun run typecheck   # tsc --noEmit
bun run test        # bun test
bun run verify      # all three

Repo layout

openseek/
├── packages/         14 workspace packages
│   ├── core/         shared leaf utilities
│   ├── provider/     25+ provider adapters (Vercel ai SDK)
│   ├── session/      main agent loop + compaction strategies
│   ├── tool/         built-in tool registry
│   ├── command/      slash-command registry
│   ├── tui/          @opentui/solid terminal renderer
│   ├── mcp/          MCP client (stdio / SSE / websocket)
│   ├── skill/        skill loader
│   ├── agent/        sub-agents + RLM fan-out
│   ├── memory/       SessionMemory
│   ├── plugin/       plugin protocol
│   ├── server/       HTTP / SSE runtime API
│   ├── lsp/          LSP client (tsserver, rust-analyzer, pyright, ...)
│   └── cli/          entrypoint
├── scripts/          init / build / lint / typecheck / test / verify
├── tests/            cross-package smoke / e2e / coverage gate
├── install/          curl-bash installer + Brewfile + Nix expression
├── bin/openseek      executable shim
└── package.json      bun workspaces root

License

See LICENSE.

Similar Articles

@ShenHuang: The number of stars an open-source project gets has nothing to do with code quality, release timing, or even features. WhaleBro's DeepSeek-TUI May 1 tweet got 375K views and 900 stars, while my similar open-source product OpenSeek got 51K views and 70 stars on its post yesterday...

X AI KOLs Following

This tweet analyzes the secret to getting stars for open-source projects, pointing out that the key lies in building identity recognition early in big accounts' comment sections, not just code quality. The author uses their OpenSeek and WhaleBro's DeepSeek-TUI as examples to illustrate the importance of reach and subsequent commercialization.

@shachepi: 天下苦沉浸式翻译久矣。 除了昨天的陪读蛙,KISS Translator 也是个顶级平替。 纯粹,完全开源。界面清爽。除了网页翻译,它同样自建接口支持非常全(Claude、Gemini 等各类AI都能接)。不想被商业插件割韭菜,用这种自接…

X AI KOLs Timeline

天下苦沉浸式翻译久矣。 除了昨天的陪读蛙,KISS Translator 也是个顶级平替。 纯粹,完全开源。界面清爽。除了网页翻译,它同样自建接口支持非常全(Claude、Gemini 等各类AI都能接)。不想被商业插件割韭菜,用这种自接 API 的最稳,加载也快,效果还好! 还没找到顺手平替的,再试试这个。 项目地址:https://github点com/fishjar/kiss-translator

@laowangbabababa: Hilarious. Big tech companies like Alibaba and ByteDance never expected their own jargon would be used to PUA AI. 18.8k stars, open source, MIT. The project is called PUA AI, packed with 14 types of big company lingo, automatically switched based on the task. If you ask AI what the underlying logic is, it goes into Alibaba closed-loop mode. If you ask AI about ROI, it switches to ByteDance data and A/B testing…

X AI KOLs Timeline

The project PUA AI collects 14 types of internal jargon from major Chinese tech companies and automatically switches between them to guide AI behavior. Experiments show it improves fix points by 36%, verification steps by 65%, and tool invocation and hidden issue discovery by 50%.