I used Claude 3 to unredact this part from the OpenAI emails. What's wild is that they used a "per word" redaction, meaning each redaction length is proportional to the length of the words, so assuming context and word length, this is Claude's guess using the page source:
Claude 3 Unredacts OpenAI Emails Using Word Length Patterns
By
–
