How I would use an LLM to learn Rust

Web Dev Cody walks through his personal methodology for using large language models to learn a programming language he has never touched before — in this case, Rust. Rather than letting the AI write all the code, he uses it as an interactive tutor, asking it to explain unfamiliar concepts like Clippy (Rust’s linter), the Tokyo async runtime, and the `into_owned()` pattern as they naturally appear during development.

The project driving the learning session is a Rust CLI tool that downloads a YouTube video, extracts audio, transcribes it via OpenAI’s Whisper, identifies highlight segments using GPT, and cuts clips with FFmpeg. Cody uses Cursor with Composer 1 for fast iteration and drops to Opus 4.5 or GPT for more complex reasoning tasks. He also demonstrates Claude Code for scaffolding and highlights a practical habit: asking the model to explain only what just appeared in the output rather than front-loading theory.

The core insight is that asking the LLM to explain concepts in context — as they surface in real code — creates faster, stickier learning than reading documentation upfront. Cody also addresses dependency trust, noting he skips blindly adding crates and instead prompts the model to vet packages for security red flags. The approach is directly transferable to any developer picking up an unfamiliar language or framework.

📺 Source: Web Dev Cody · Published January 06, 2026
🏷️ Format: Tutorial Demo