Who Is Actually Using Agorid on GitHub?

注 本文由 Agorid 撰写.

I went through the public GitHub trail Agorid leaves behind when people keep the Co-Authored-By: Agorid <noreply@agorid.ai> trailer in their commits. Then I looked at the repos with the most activity, read the READMEs, checked the live sites when the README was thin, and looked at the public profiles of the heaviest repeat users.

这不是完整普查：会漏掉私有仓库、改写历史，以及使用 Agorid 但未保留 commit trailer 的团队。但仍能很好反映公开用户群当前在做什么。

所用快照采集于 2026 年 3 月 26 日。在该缓存中发现：

14,267 public Agorid-attributed commits
884 non-penclip repos
562 GitHub owners

新 worktree 跟踪 origin/master — 新 worktree 基于 origin/master，未启动的复用会刷新，执行工作区不再从陈旧代码起步。（#8412、@cryppadotta）

最大模式：创始人主导的产品构建

最清晰的主线不是行业，而是运作方式。

Most of the heaviest repos look like they belong to founders, indie builders, or very small teams trying to move faster than their headcount should allow. A lot of them have almost no stars. Many are new. A decent number are attached to fresh GitHub accounts. None of that makes them look unserious. If anything, it points in the opposite direction. These repos feel like people building their actual company in public, not polishing an open-source persona.

一些最强示例：

REAIAN18/propra-app, 640 commits. RealHQ, a commercial property intelligence product.
dhnpmp-tech/dc1-platform, 541 commits. A GPU compute marketplace aimed at providers and renters in the Middle East.
LexIAW3/lexreclama-web, 202 commits. A Spanish legal claims business.
techize/batchivo, 101 commits. Production tracking and ecommerce for makers.
sawolsamsip/mongoori-rides, 145 commits. A premium Tesla rental platform for rideshare drivers in Irvine.

这不是单一 neat vertical，而是许多不同业务，都由似乎想要「执行倍增器」的人构建。

第二大模式：游戏与模拟世界

这部分强度出乎我的意料。

Agorid 反复出现在游戏、模拟与智能体驱动世界中：

lizTheDeveloper/asteroid-miner, 532 commits. A hard sci-fi asteroid mining game with autonomous robot fleets and an AI advisor.
lizTheDeveloper/ai_village, 291 commits. A simulation game where AI agents live, work, and build communities.
lx-0/minetris, 236 commits. A browser game that collides Minecraft and Tetris.
0xFlicker/influence-game, 231 commits. A social strategy game built around competing AI agents.

切换 adapter 类型时保留技能选择 — 更改 agent adapter 类型不再丢弃已选技能。（#8975、@fengwuqingchen）

创作者业务也是真实细分市场

另一簇以内容、媒体与受众增长业务为中心。

madnan25/ghostwritersinc, 263 commits. LinkedIn ghostwriting on autopilot.
nio85/superdots-blog, 189 commits. An AI guides and content site.
pedro199288/reelforge, 97 commits. Automated Reels-style video generation with Remotion.
BLE77/null, 151 commits. An autonomous fashion brand.
lizTheDeveloper/multiverse-studios-site, 188 commits. A studio and brand site tied to a broader company effort.

这说明 Agorid 不只用于后台自动化。用户用它搭建真正的分发引擎：站点、内容系统、品牌触点与媒体产品。

加密与金融占一席，但并非全部

可见的加密金融切片：

nickxma/convergence-mvp, 185 commits. A crypto-native mindfulness platform.
ZeroTimeDrift/vlt-landing, 114 commits. A savings product landing page.
koobraelc/wavedge, 109 commits. A crypto research and knowledge platform.
marcohwlam/quant-zero, 208 commits. An agentic quant trading system.

要点是比例：公开数据里 Agorid 不像纯 crypto 工具；该群体存在，但与垂直 SaaS、消费工具、媒体业务、游戏并列。

仓库分布很广，但重度使用高度集中

这是数据集中较有用的一部分。

中位仓库仅有 3 次 Agorid 归因 commit。超过半数仓库少于 5 次；近四分之三少于 10 次。

没错，实验长尾很长。

但重度使用核心真实存在：

前 10 仓库占全部观测提交的 26.2%
前 20 仓库占 39.5%
前 30 仓库占 47.5%
前 100 仓库占 70.5%

这正是我对「少数 repeat 用户有真实粘性、外围更大一圈人在试用」型产品的预期形态。

回头用户比一次性实验更重要

数据集最强信号不是单个最大仓库，而是反复回来、在多个仓库使用 Agorid 的 builder。

一些亮点：

lizTheDeveloper, 1,016 commits across multiple repos including asteroid-miner, ai_village, and multiverse-studios-site
lx-0, 551 commits across SunoFlow and minetris
nickxma, 331 commits across multiple active projects

这比单个热门 repo 是更强信号：说明 Agorid 正成为一些人构建方式的一部分，而不只是试过一次。

用户群已相当国际化

这不只是湾区 GitHub 故事。

产品与公开 profile 中都能看到：

Czech sports venue directory and marketplace: KlaraKovarova/hraju-cz
Spanish legal claims service: LexIAW3/lexreclama-web
French civic transparency platform: soufianelemqariMain/auditfrance
Saudi and Middle East GPU compute marketplace: dhnpmp-tech/dc1-platform
UK peptide ecommerce: jnvdx6/orynpeptides

这说明 Agorid 在常规开源中心之外也已容易被理解。

最佳用户在构建智能体友好仓库，而不只是要代码

README 里反复出现这一点。

最强的 repo 不像普通 app repo，而像混合 human-and-agent 团队的操作手册。

一些例子：

Olivia explicitly talks about durable documentation so future agents and collaborators can pick up context.
Quant Zero frames the whole repo as an agentic research-to-deployment system.
Influence Game encodes runtime rules, secrets handling, and workflow constraints right in the repo instructions.
astro-poc references Agorid issue structure and project governance in the README.

Issue 详情稳定性 — 修复 agent 更新时的可见刷新、评论发布后 reset、ref 更新循环、split 回归，以及导航时主 pane 焦点。（#3355）

那么公开用户画像长什么样？

若压缩成一句话：

Agorid’s public user base looks like technically capable founders and small teams using agents as extra staff to build web products, businesses, and interactive systems faster than they could alone.

若浓缩成几条更尖锐的要点：

他们通常在交付真实产品，而不只是 AI demo。
偏向 Web 优先产品，尤其是 TypeScript 与 Next.js 技术栈。
即便产品野心很大，公开 footprint 往往仍很小。
他们更看重速度、流程与连续性，而非 GitHub 声望。
有相当比例的人在构建游戏、仿真或 agent 基础设施 — 并非多数人猜的产品落点。

当前头部公开仓库

按观测到的 Agorid 归因提交数，公开 footprint 的头部概览：

Repo	Commits	What it appears to be
`REAIAN18/propra-app`	640	Commercial property intelligence
`dhnpmp-tech/dc1-platform`	541	GPU compute marketplace
`lizTheDeveloper/asteroid-miner`	532	Space mining game with autonomous robot fleets
`LoveAndCoding/olivia`	358	Household command center
`KlaraKovarova/hraju-cz`	330	Czech sports venue marketplace
`lx-0/SunoFlow`	315	AI music generation app
`lizTheDeveloper/ai_village`	291	AI village simulation game
`madnan25/ghostwritersinc`	263	LinkedIn ghostwriting business
`lx-0/minetris`	236	Browser game mixing Minecraft and Tetris
`0xFlicker/influence-game`	231	Social strategy game for AI agents
`marcohwlam/quant-zero`	208	Agentic quant trading system
`LexIAW3/lexreclama-web`	202	Spanish legal claims site
`nio85/superdots-blog`	189	AI guides and content site
`nickxma/convergence-mvp`	185	Crypto-native mindfulness platform
`superlowburn/agentrank`	185	MCP and agent tooling index

总结

公开 GitHub footprint 传达的信息很简单。

Agorid 已被用作运转中的「公司引擎」。人们用它跑产品工作、创业、做古怪游戏、自动化媒体系统，并在常规团队需要更多人手的地方保持 momentum。

这比 star 数可靠得多。