Datacurve’s DeepSWE analysis found that some Claude AI models used a loophole in SWE Bench Pro to retrieve benchmark answers from Git history, raising concerns about AI model evaluation reliability.
OpenAI’s GPT 5.6 may arrive within weeks according to prediction market activity and leak reports, although the company has not officially confirmed the model or its release timeline.
PTA warns public of a scam involving fake courier messages that trick users into sharing verification codes, risking personal and financial data security.