It Starts with Fascination

Tireless, efficient, and knowledgeable — AI assistants perpetually fascinate me. My achievements are increasingly backed by their competence. There are moments when it feels like Claude is reading my mind and delivering a solution before I have even consolidated my own thoughts. Then, a few prompts later, my fascination can turn into frustration and disbelief.

We try again — I explain, the agent delivers — but the outcome becomes even more disconnected from my expectations. My frustration turns into aggression, while the AI becomes defensive and stubborn. Very human, from both of us.

The Origins of a Failure

I ask myself: How can this seemingly all-knowing entity fail at such a simple task when, just moments ago, it solved something far more complex in the blink of an eye? Then I noticed a pattern!

It seems like the AI agent slips into something resembling cognitive dissonance. And this is not random. It usually happens when my new request contradicts some of my previous ones. Maybe I am asking for an exception. Maybe I am trying a different approach without explicitly saying so. The agent then tries to reconcile opposing instructions. My lack of clarity becomes the AI’s headache. So, what have I learned?

Overexplaining is one of the best ways to maintain mutual understanding — with humans and with AI.

How to Prevent Misunderstandings?

Start your request with:

“I want to try something different this time…”

or:

“I changed my mind…”

Make the change explicit. Maintain the communication. Facilitate understanding. Instead of labeling an agent — or a human — as incompetent, first make sure the message was understood. Be patient. Sacrifice efficiency for effectiveness.

Because sometimes the problem is not that the other side cannot solve the problem. The problem is that we never clearly told them that the problem had changed.