WHY DOES OPUS 5 FEEL WORSE TO WORK WITH?

Summary from the TLDR Dev newsletter

Working with Opus 5 feels less effective compared to its predecessors primarily because it requires more oversight and fails to seek clarification in ambiguous situations. This issue comes from a focus on developing self-improving AI systems that excel on benchmarks, which often focus on bold assumptions over the necessary inquiry and flexibility required in real-life tasks.