OpenAI shares GPT-6 Astra prompting tips to curb unwanted behavior

openai shares tips for gpt-6 astra, including prompting the model to infer user intent and show a 'bias towards action'. gpt-6 astra asks clarifying questions more often than gpt-5.6 sol and tends to structure responses with lists and tables. openai recommends auditing skill files and context documents, and using a blocklist of 'slop words' like 'delve', 'foster', and 'leverage' to shape the model's writing style. this affects gpt-6 astra users and ai developers who want to work around the model's tendencies.