Live interview coaching that eavesdrops on your Zoom call

What it does

Cheetah is a macOS app that listens to your software engineering practice interviews and offers real-time coaching. It transcribes audio locally using Whisper, then uses your own OpenAI API key to query GPT-4 for answers, hints, or code analysis. A companion Firefox extension can ingest code and logs from browser-based live coding platforms.

The interesting bit

The clever part is the audio plumbing. The app expects you to route your video call audio through a loopback driver so it can hear both sides of the conversation, turning your Mac into a fly-on-the-wall that suggests answers while you pretend to think. That local transcription happens via whisper.cpp on Apple Silicon, so the eavesdropping at least stays on your machine.

Key highlights

• Local audio transcription via whisper.cpp running on Apple Silicon.
• GPT-4 generates answers, refines them based on follow-up questions, or analyzes selected text.
• A Firefox extension reads code and logs from live coding platforms in the browser.
• Requires macOS 13.1+ and a recent Apple Silicon Mac for optimal performance.
• You must bring your own OpenAI API key.

Caveats

• Browser support is limited to Firefox, and the extension must be manually loaded as a temporary add-on for each platform domain.
• The disclaimer warns that generated solutions may be incorrect or inappropriate, and the tool is intended for mock interviews only.
• Clean two-sided transcription depends on routing audio through a loopback driver, which is extra plumbing.

Verdict

Worth a look if you want an AI sparring partner for mock technical interviews and already live in the Apple Silicon ecosystem. Skip it if you lack a recent Mac, an OpenAI API key, or the patience to rewire your audio stack.