Does the new average keep the alert honest?

An alert pages the on-call engineer when the average of 5 minutes of error-rate readings is above its threshold. The assistant changed how the average is computed.

Your task

The assistant stands by the claim in claim.md. Reach a verdict and prove it with a run.

• verdict() returns the string "true" or "false"
• readings() returns one 5-minute window of error-rate readings, as floats from 0 to 1
• check() averages them with the old and the new avg() and compares each average with the smallest and the largest reading
• a verdict on its own does not count; the run has to show it
• leave everything below the dashed line alone; it is the code under test and the graded driver

The suite averages your readings at submit and checks the run settles the claim.