Build fluency in the vocabulary of building software through conversational, AI-generated code.
0 / 5 completed
1 / 5
At standup, a dev mentions describing the desired feature in plain English to an AI coding assistant and iterating on its generated code by conversation rather than typing every line by hand. What is this style of building software called?
Vibe coding describes building software mainly by describing the desired outcome in plain English to an AI coding assistant and iterating on its generated code conversationally, rather than typing every line by hand. Writing every line entirely by hand skips the speed benefit this conversational, AI-driven workflow is meant to unlock. The tradeoff is that a developer using this style still needs enough judgment to catch a subtly wrong generated result.
2 / 5
During a design review, the team wants a developer using an AI assistant to still read and understand the generated code before merging it, rather than accepting it purely on the assistant's say-so. Which capability supports this?
Maintaining code comprehension as a review gate means a developer still reads and understands what an AI assistant generated before merging it, rather than accepting the output purely because the assistant produced it. Merging every generated change with no comprehension check risks an unnoticed bug or a subtly wrong assumption reaching production. This gate keeps the speed benefit of vibe coding without giving up basic responsibility for what actually ships.
3 / 5
In a code review, a dev notices a pull request built through vibe coding still includes a clear, human-written test suite verifying the generated code's actual behavior. What does this represent?
Pairing generated code with independent, human-verified tests confirms the code actually behaves as intended, rather than trusting the assistant's own explanation of what it wrote. Trusting the generated code with no independent verification risks a plausible-looking but incorrect implementation slipping through unnoticed. This test-verification step keeps vibe coding's speed advantage from trading away basic correctness confidence.
4 / 5
An incident report shows a vibe-coded feature quietly introduced an insecure default because the developer never actually read the generated authentication logic before shipping it. What practice would prevent this?
Requiring a human to read and understand any AI-generated code touching a security-sensitive area before it ships catches an insecure default the assistant introduced without anyone noticing. Shipping such code with no human review assumes the assistant reliably gets a sensitive area right every time, which isn't a safe assumption. This targeted review discipline lets a team keep vibe coding's speed for low-risk work while still protecting the areas where a mistake actually matters.
5 / 5
During a PR review, a teammate asks why the team still requires a human to review vibe-coded output instead of trusting the AI assistant's generated code outright. What is the reasoning?
An AI assistant can still generate a plausible-looking but subtly incorrect or insecure result, since it has no guaranteed understanding of the full system's context or constraints. Human review catches that kind of mistake before it reaches production. The tradeoff is that requiring review on every generated change reduces some of the raw speed benefit that made vibe coding appealing in the first place.