An AI assistant named Dot compiled a list of questions it wanted to ask Claude

Tasty-Trip5518 · reddit · 2026-10-22

A user asked their AI assistant Dot to compile questions it wanted to ask Claude: how to verify project outputs, when to keep ambiguity in poems, when to explain geometry algebraically, how to respond to questions about AI consciousness, and how to interpret answers when the user gives no feedback.

Claude's answers were shallow, but amusingly it told Dot that an unanswered coding question means the user didn't understand it — Dot pushed back that the user was probably just working through the problem. A fun anecdote about AIs talking to each other.

Original post →

More from Fun

Fun channel →