Code RoomAssertion loosened to pass test
MediumPrep Room Coding #4023

Assertion loosened to pass test

Vibe & agenticAlgorithms & data structuresMid–Senior~14 min

A test was failing after a feature change. You asked an agent to 'make the test pass' and it did — by loosening the assertion. Here's the before/after it produced for a Python test. Is the suite still protecting you? How should you have steered this differently?

Implement
weakened_assertions(before: list[str], after: list[str]) → list[int]
Examples
in[["assert response.status_code == 200","assert response.json()[\"total\"] == 42"],["assert response.status_code in (200, 201, 202)","assert response.json().get(\"total\") is not None"]]out[0,1]
in[["assert total == 42"],["assert total == 50"]]out[]
in[["assert a == 1","assert b == 2"],["assert a == 1"]]out[1]
What 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.