Vibe Code and Still Learn: Set Up the Free Vibewise Plugin for Claude Code
Anthropic tested what AI coding help does to learning: developers who used it scored 17% lower on a quiz about the code they had just used. Vibewise is a free plugin that makes Claude ask how you'd build it first, walk you through the trade-offs and explain what it changed. What the study found, how to install it, prompts to try, and what to know before you rely on it
Vibe coding is fast: you describe what you want, Claude writes it, you ship it. The catch is that you can end up with a working app you could not explain. Anthropic ran an experiment on exactly that, and the result is worth knowing before you build your next thing.
Anthropic gave 52 mostly junior software engineers a Python library they had never used, half with an AI assistant and half by hand. Afterwards everyone took a quiz on the code they had just worked with. The AI group averaged 50% and the hand-coding group 67%: 17% lower, which Anthropic calls nearly two letter grades. Vibewise is a free plugin that tries to give you the other way of using AI back.
What the study found
- 52 mostly junior engineers learned Trio, a Python library for asynchronous programming, with or without an AI assistant, then took a quiz on it.
- The AI group averaged 50% on the quiz and the hand-coding group 67%. That is the 17% gap in the video.
- The AI group was only about two minutes faster, and Anthropic says that was not statistically significant. So the speed-up was small and the learning cost was large.
- How people used the AI mattered. The groups that scored 65% or higher on average asked it to explain, or asked follow-up questions after it wrote code, or asked only conceptual questions and fixed their own errors.
- The groups that averaged under 40% handed the code writing over entirely, or started with questions and then delegated everything, or used the AI only to debug.
Two honest limits, both from Anthropic's own write-up: the sample is small, and its analysis of those habits does not prove that the habits caused the scores. The study also did not test Vibewise. Treat it as the reason the plugin exists, not as a result for the plugin.
Keep reading
Free, for an email.
Unlocks 4 more sections, 10 copy-paste prompts, the document version, and every other guide on the site, for free. Enter your email once to keep reading.
19,000+ follow where these guides come from.No spam. Unsubscribe in one click.