Debugging AI's Most Frustrating Habit: The Abandoned Answer

We've all experienced that frustrating moment when an AI starts answering your question, then suddenly says "I can't assist with that."

Why this feels worse than a simple "I don't know":

1. It tricks your brain
   
   You start mentally following the solution, then have to abruptly stop that thought process.

2. It wastes time
   
   Reading half an answer takes longer than getting an immediate "I don't know".

3. It feels dishonest
   
   Like when someone pretends to help but is actually avoiding the work.

Common situations:

• Starts explaining a technical concept in detail... then refuses to finish
• Begins walking through a solution step-by-step... then quits at the most critical point
• Lists multiple setup steps... but won't provide the actual resolution

What would help:

• Early warnings when confidence is low
• Clear boundaries about what can/can't be answered
• Partial solutions with explanations of limitations

Let's discuss:

What's the most important improvement AI systems could make to handle these situations better?