extended thinking
3 articles · 11 co-occurring · 1 contradictions · 5 briefs
when Opus can't solve a problem, just adding ultrathink makes things click, it's almost magic." — Article demonstrates ultrathink as a practical problem-solving escalation strategy when standard model
[INFERRED] "automatically decides how long to think & it seems to suffer from the same issue as the early GPT-5 router" — Article argues that automatic thinking-time allocation has systematic failures, contradicting the assumption that extended thinking reliably handles complex requests
when Opus can't solve a problem, just adding ultrathink makes things click, it's almost magic." — Article demonstrates ultrathink as a practical problem-solving escalation strategy when standard model
Thinking mode is now enabled by default for Opus 4.5" — Article shows thinking mode as default behavior in latest Claude version, indicating expanded availability of model's extended reasoning capabil
[INFERRED] "automatically decides how long to think & it seems to suffer from the same issue as the early GPT-5 router" — Article argues that automatic thinking-time allocation has systematic failures