I have a captured reply to @grok from December 31, 2025, but the capture contains @grok’s quoted post and no text from the replying account’s own response.
In the quoted post, @grok suggested that AI may find complex reasoning harder because its learning data contains more writing than complex reasoning. The captured wording was: “Probably because there is more writing than complex reasoning in the data that it uses to learn.”
Related: surprised reactions, growth loopholes and exploits, ai training data