I spent last weekend digging through GitHub like a kid in a toy store—except instead of toys, I found tools that make me wonder why nobody told me about them sooner. Here’s a handful that genuinely changed how I do things, from turning old phones into microphones to decoding USB-C cables.
01 Turn Your Old Phone into a Wireless Mic
You know that drawer full of retired Androids? MicYou turns one into a wireless microphone for your computer. It connects over Wi-Fi, USB, or Bluetooth—Wi-Fi being the laziest (phone and PC on the same network, done), USB via ADB for lowest latency, and Bluetooth as a fallback when you’re off-grid.
The audio processing built in: noise suppression, auto gain, echo cancellation. No extra software. The desktop client runs on Windows, Linux, and macOS; the phone side is Android. You can tweak sample rate, channel count, and audio format. The whole thing is written in Kotlin Multiplatform, so Android and desktop share a codebase.
I tried it for a last-minute Zoom call when my headset died. Grabbed my old Pixel 3, installed the app, connected via Wi-Fi—took maybe two minutes. The audio quality surprised me; it’s cleaner than most Bluetooth mics I own. Only catch: you need both devices on the same network for Wi-Fi mode, but USB mode works offline.
02 Finally Understand What That USB-C Cable Can Do
We’ve all got that pile of USB-C cables where half are mystery meat. WhatCable is a macOS menu bar tool that reads the e-marker chip inside the cable and tells you—in plain English—exactly what it supports: Thunderbolt 5, USB4, USB 3.0, or just charging.
It also diagnoses slow charging. Is it the cable’s fault or the charger’s? The tool shows all voltage profiles your charger supports, highlights the one currently negotiated, and lists the cable’s actual speed rating and chip vendor.
I blamed my MacBook Air’s slow charging on a cheap Anker cable for weeks. Turns out the cable was fine—my charging brick was the bottleneck. WhatCable saved me $20 on a new cable I didn’t need. It only works on Apple Silicon Macs (M1 and up) because Intel Macs don’t expose Thunderbolt controller data. Install via Homebrew:
brew tap darrylmorley/whatcable && brew install --cask whatcable 03 Command-Line Office Document Manipulation (No Office Installed)
Python’s python-docx and openpyxl are decent but fall apart on complex layouts. LibreOffice’s headless mode works but is a 400MB hammer. OfficeCLI is a single-file C# binary that needs no .NET runtime, no Office installation—just download and run. It creates, reads, and modifies Word, Excel, and PPT files from the command line, outputting structured JSON.
The architecture has three layers: semantic reading (ask “what’s in this doc?”), DOM operations (modify specific elements by path), and raw XML editing via XPath. There’s even a watch mode that shows changes live in a browser.
I used it to automate a weekly report generation: pull data from an API, fill an Excel template, convert to PDF. The pipeline took 30 seconds what used to be a 20-minute manual chore. For AI agents that need to produce Office documents, OfficeCLI is a gift.
04 AI That Writes Your Software Copyright Application (China-specific)
If you develop software in China, you’ll need a software copyright registration for app stores, government bids, or compliance. The paperwork is tedious: a user manual, source code excerpts, and an application form, all requiring strict formatting. A manual run takes half a day.
SoftwareCopyright-Skill is a skill for Codex that reads your project source code and auto-generates all three documents. The user manual describes your actual business logic—not a cookie-cutter template. The source code excerpt picks real code (first 30 pages, last 30 pages, as required). It doesn’t fabricate anything.
The process has six checkpoints: business context, form fields, code selection, screenshots, and final review. You can adjust at each step. Outputs: a .docx user manual, a .docx source code document, and a .txt with info for the registration website.
I tested it on a small side project. It cut the preparation time from four hours to 45 minutes. The generated manual actually made sense, which surprised me because most autogenerated ones are gibberish.
05 An AI Programming Jargon Buster
Matt Pocock’s Dictionary of AI Coding is a plain‑English glossary for terms like context window, token vs. turn, sycophancy, hallucination, vibe coding, and grilling. It’s organized into seven chapters covering model basics, sessions & context, tools & environments, failure modes, handoff patterns, memory & prompting, and working styles.
Each term comes with a real‑world dialogue example and a note on common confusions—like attention degradation (forgetting due to long context) vs. hallucination (making things up). Pocock even coins terms like smart zone and dumb zone to describe how an agent’s performance degrades as context grows.
I’ve found myself referencing it when debugging weird LLM behavior. For anyone pair‑programming with AI, bookmark this repo. It’s pure Markdown, readable on GitHub, no install needed.
06 Animal Crossing‑Style React UI Components
Last one is pure fun. animal-island-ui is a React component library styled like Nintendo’s Animal Crossing: cute, rounded buttons, cards, switches, time displays, even a phone UI. The author built a personal website template using these components, plus a children’s English‑learning app called HiKid.
I’m planning a small game for my niece, and this library saves me the design effort. It’s lightweight and easy to drop into a React project. If you want to make something playful—a blog, a quiz, a kid‑friendly tool—this is a delightful shortcut.