Observability in software development should prioritize error tracking over traditional logs, metrics, and traces, as exceptions provide the clearest indication of failures in the code. By focusing on capturing detailed context around errors, developers can gain invaluable insights that are often lost in the noise of standard observability practices. The author argues that the current approach to observability tends to downplay the importance of errors, which should be treated as first-class signals when diagnosing issues.