Claude’s Code Quality Conundrum Continues

Source: DevOps.com

Anthropic admits to a month‑long degradation in Claude’s output due to reasoning “effort” trade‑offs, cache bugs, and verbosity prompts. As Opus 4.7 rolls out with mixed developer reviews, the company is doubling down on “frontier safety” with the restricted Mythos model while leaning on high‑profile advisors like Rishi Sunak to navigate global AI regulation.