Inefficient Docker layer caching
An AI wrote this Dockerfile for a Go service and your CI builds went from 40s to 6 minutes after a one-line app change. The build is correct — but something about layer caching is broken. What did the AI do wrong?
Implement
rebuilt_step_count(steps: list[str], changed_files: list[str]) → intExamples
in
[["copy|.","run|go mod download","run|go build"],["cmd/server/main.go"]]out3in
[["copy|go.mod go.sum","run|go mod download","copy|.","run|go build"],["cmd/server/main.go"]]out2in
[["copy|go.mod go.sum","run|go mod download","copy|.","run|go build"],[]]out0What a strong answer looks like
Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
0:00 of about 14 min
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.