- 开发者使用搭载M5芯片的128GB MacBook Pro进行本地大语言模型测试,发现其性能表现优异。因对系统自带活动监视器不满,采用“vibe coding”方式开发了两款macOS菜单栏工具:Bandwidther用于监控各应用网络带宽使用情况,Gpuer用于显示GPU运行状态。两款应用均基于SwiftUI开发,代码可完整写入单个文本文件,无需打开Xcode即可通过Claude Opus 4.6或GPT-5.4生成。
Bandwidther可实时展示总下载/上传速度、60秒带宽趋势图、累计数据量、连接数及目标地址,并按进程列出如Dropbox、nsurlsessiond等应用的详细流量数据。该工具以菜单栏图标形式运行,点击弹出信息面板。
M5 MacBook Pro本地LLM性能强
vibe coding开发macOS工具
Bandwidther监控网络流量
菜单栏工具无需Xcode
- Developer Creates macOS Performance Monitoring Tools Using AI-Assisted "Vibe Coding"
A developer has built two lightweight macOS applications—Bandwidther and Gpuer—using AI-assisted "vibe coding," a method where large language models (LLMs) generate functional code with minimal manual intervention. The tools were developed for a 128GB M5 MacBook Pro to monitor system performance, specifically network and GPU usage, after dissatisfaction with Apple’s built-in Activity Monitor. Bandwidther tracks real-time network bandwidth by process, displaying download/upload speeds, a 60-second bandwidth graph, cumulative totals, and per-application usage—initially created to investigate Dropbox’s network activity. Gpuer provides GPU utilization insights. Both apps are implemented as menu bar utilities with expandable panels for detailed metrics. The developer used Claude Opus 4.6 and GPT-5.4 to generate full SwiftUI applications that fit within a single text file, enabling development without opening Xcode. This marks the second macOS app created via vibe coding, following a prior presentation app. The approach demonstrates the growing capability of advanced LLMs in generating production-ready, platform-specific GUI applications with minimal human input. The apps are open source, with Bandwidther’s code shared via a public gist.
Key Takeaways:
AI models can generate functional SwiftUI apps without Xcode
Vibe coding enables rapid development of system monitoring tools
Bandwidther provides detailed per-process network usage insights
M5 MacBook Pro supports efficient local LLM development workflows
Source: Original Article
查看原文 →
View Original →