================================================================
Two "Awesome" AI Collections, Often Confused: A Source-Checked Inventory of the 132-App Repo
两个「Awesome」合集别混淆：132 个 AI 应用合集核源盘点
----------------------------------------------------------------
vertciti Data Room · Public Document
Date: 2026-10-20 | Version: v1
Sources: GitHub API directory counts (checked 2026-10-19); repositories
  Arindam200/awesome-ai-apps and Shubhamsaboo/awesome-llm-apps.
  See source notes inline.
Classification: Public — contains no confidential information.
================================================================

----------------------------------------------------------------
ENGLISH
----------------------------------------------------------------

Introduction

A widely shared video presents a collection of "132 AI applications" — but its visuals and its narration point at two different GitHub repositories. On screen is Arindam200/awesome-ai-apps (roughly 16,000 stars); the name spoken aloud, "Awesome LLM Apps," belongs to a different repository with roughly 142,000 stars, Shubhamsaboo/awesome-llm-apps. Our fixed order for checking any hyped collection: untangle the identity first, then judge the contents. (Figures as checked on 2026-10-19.)

Source-checked points

1. Identity, verified. The on-screen repo's README describes 132 projects; counting the actual project directories through the GitHub API, category by category, gives 137 (Starter 21, Simple 20, Advanced 35, RAG 18, MCP 14, Memory 13, Voice 9, Fine-tuning 6, Course 1) — the same magnitude. The repo is MIT-licensed, created in February 2025, still actively updated, and sponsored by Nebius Token Factory.
2. Structure. Of the nine categories, advanced applications are the largest group at 35; the starter section is notable for breadth, with one getting-started template for each of 21 frameworks.
3. Code weight. In our sampling, over sixty percent of the projects are single-file demos. A minority are full multi-file applications — a financial-document processing pipeline, one of the repo's rare test-carrying projects (an invoice auditor), and MCP security tooling that gates data access by least privilege. Treat this repository as a library of ideas and skeletons; anything headed for production needs per-project verification.
4. The real cost of "just run it." The steps are genuinely short — clone, install dependencies, add a key, run (Python 3.10+) — but nearly every project needs an API key and defaults to the sponsor's endpoints. The saving grace is that access goes through an OpenAI-compatible interface, so pointing a project at your own channel is a base-URL-and-key change.
5. Two circulating claims fail source-checking. "No. 2 on GitHub's daily trending chart": GitHub keeps no official archive of daily trending ranks, so there is nothing to verify it against. "Spanning 20+ industries": the repo has no industry classification at all, and its identifiable vertical spread is roughly 8–10 areas. Neither claim is used here.

The vertciti view

Our standing rule for open-source agent assets is named mining, one project at a time: pick only the few projects that match a line we are actively building, repoint them at our own channels, run them for real, and write a one-page verification before anything counts as acquired. The category bookmark we keep is awesome-llm-apps; from the Arindam200 collection we take named items only — MCP security, social research, invoice auditing — and never import the whole repo.

Sources

- The on-screen repo: https://github.com/Arindam200/awesome-ai-apps
- The spoken-name repo: https://github.com/Shubhamsaboo/awesome-llm-apps

----------------------------------------------------------------
中文
----------------------------------------------------------------

导语

社交媒体上流传着一个「132 款 AI 应用合集」，画面与口播其实指向两个不同的 GitHub 仓库：画面里的是 Arindam200/awesome-ai-apps（约 1.6 万 star)，口播说的「Awesome LLM Apps」则是另一个约 14 万 star 的仓库 Shubhamsaboo/awesome-llm-apps。先分清真身，再看内容成色——这是我们核查任何热门合集的固定顺序。

核源要点

1. 真身核对：截图仓库的 README 自述收录 132 个项目；我们经 GitHub API 逐类点算，实际项目目录共 137 个（入门 21／简单应用 20／高级应用 35／RAG 18／MCP 14／记忆 13／语音 9／微调 6／课程 1)，与自述同量级。该仓 MIT 许可，2025 年 2 月创建，至今仍在活跃更新，由 Nebius Token Factory 赞助。（数字为 2026-10-19 核查时点）
2. 内容结构：九大类中高级应用最多（35 个）；框架入门模板覆盖面宽，21 个框架各有一份起步模板。
3. 代码成色：抽查显示六成以上是单文件演示，少数才是多文件成型应用——如金融文档处理流水线、全仓少有的带测试的发票审计工具、按最小权限管控数据访问的 MCP 安全工具。把这个仓库当「灵感与骨架库」是合适的；当「拿来即用的生产模板库」，则每一件都需要单独验证。
4. 「拿来就能跑」的真实门槛：流程本身确实短（克隆 → 装依赖 → 填密钥 → 运行，Python 3.10+)，但几乎每件都要 API key，且默认接入赞助商的服务。好在接入走的是 OpenAI 兼容接口，换成自有通道只需改接入地址与密钥。
5. 两项网传说法不成立：「GitHub 趋势日榜第二」——GitHub 官方不存档每日趋势名次，无源可核；「横跨 20 多个行业」——仓库并无行业分类口径，实物可辨认的垂直领域约 8–10 个。两项均经核源不成立，本篇不予采用。

vertciti 视角

我们评估开源 agent 资产执行一条固定纪律：点名挖矿、一件一验——只挑与在建业务线同题的少数项目，改指自有通道真跑通、写验收页，才算挖到。品类主书签我们留给 14 万 star 的 awesome-llm-apps;Arindam200 仓只按 MCP 安全、社媒调研、发票审计等点名件取用，不整仓收藏。

来源

- 截图实物仓：https://github.com/Arindam200/awesome-ai-apps
- 口播名仓：https://github.com/Shubhamsaboo/awesome-llm-apps

---