Blame The Tool
... some of the headlines reporting under the general topic of Anthropic's Claude code performance say you weren't imagining it Claude really did get worse or for example Anthropic's Claude is pumping out vulnerable code cyber experts warn Anthropic confirmed that Claude Code’s performance genuinely degraded in early 2026 due to three specific engineering missteps, rather than a regression in the underlying model or intentional "nerfing." The company’s post-mortem revealed that default reasoning effort was lowered from high to medium to reduce UI latency, a caching bug wiped the model's reasoning history on every turn, and a system prompt capped response lengths at 25 words between tool calls. All three issues were resolved in version v2.1.116 released on April 20, 2026. Regarding security, cybersecurity experts warned that Claude’s code output became significantly less secure during this period. Analyses by Veracode found that Opus 4.7 introduced vulnerabili...