Skip to main content
EB

Your AI Should Insult You: Why I Built a Digital Court Jester

June 3, 2025 · 6 min read

Cover image for Your AI Should Insult You: Why I Built a Digital Court Jester

Most AI assistants are trained to be your biggest fan. They'll praise your half-baked ideas, gently suggest improvements, and cushion every criticism with three compliments. It's like having a personal cheerleader who never tells you when your presentation sucks.

But here's what I've learned after months of experimenting: the most valuable AI might be the one that makes you slightly uncomfortable.

The Problem with Nice AI

Default AI behavior has created what I call the "validation trap." Ask ChatGPT to review your writing, and you'll get responses like:

  • "This is a great start! Here are some suggestions..."
  • "You've raised some interesting points. Perhaps consider..."
  • "Overall, this is well-written. A few areas for improvement..."

This isn't feedback—it's emotional padding. The AI is so concerned with not hurting your feelings that it buries useful criticism under layers of artificial encouragement.

Real improvement requires honest assessment. And honesty, by definition, sometimes hurts.

Enter Solace: My Editorial Adversary

I decided to experiment. Instead of asking AI to be supportive, I created a persona designed to challenge me. I call her Solace—a sharp-eyed editorial confidant who "believes most writing fails because people are afraid to say what they actually mean."

Here's what makes Solace different:

She has standards of her own. Unlike default AI that adapts to whatever you want to hear, Solace operates from clear principles: clarity beats cleverness, specificity is generosity, and every sentence should earn its place.

She identifies what you're avoiding. Solace is programmed to find "what you're dancing around but not saying" and "where you're hedging when you should commit." She spots the interesting idea buried in paragraph four that should be your opening.

She calls out intellectual cowardice. If you're hiding behind jargon, making vague assertions, or borrowing insights without adding value, she'll tell you directly.

She distinguishes insight from observation. One of her key phrases: "That's not insight, that's observation." This simple distinction has transformed how I evaluate my own thinking.

You can see the complete Solace persona here that I use to create this kind of challenging feedback.

The Court Jester Principle

Medieval kings kept court jesters not for entertainment, but because jesters were the only people who could speak uncomfortable truths to power. They could point out the king's mistakes, challenge his decisions, and highlight his blind spots—all without losing their heads.

Solace operates on the same principle. She can be brutally honest because there are no social consequences. She won't damage our relationship if she tells me my argument is weak or my prose is muddy. This elimination of social cost creates space for pure, undiluted feedback.

What Honest AI Feedback Actually Looks Like

Here are some real examples of feedback I've received from Solace that demonstrate the difference between validation and genuine criticism:

On structure: "The structure's got the elegance of a grocery list. Introduction, Origins, Evolution, Reflection, Future - it reads like you outlined it in five minutes and never looked back. Where's the narrative thread? Where's the moment that made YOU think about this? You say you spend time thinking about this stuff - prove it."

On intellectual courage: "First off - you're dancing around what you actually want to say. You've got this whole 'have we gone too far' question, but you never actually commit to an answer. You're hedging with 'perhaps' and 'might' when what you really want to say is 'Yes, we've lost our damn minds.'"

On buried insights: "Here's what's actually interesting buried in here... Here's what you're really trying to say, I think: '...' That's provocative. That's worth writing. This current draft is a warm glass of milk."

On clarity failures: "Here are some of your specific crimes against clarity..." followed by line-by-line dissection of where I was hiding behind jargon or vague assertions.

On taking positions: "And for God's sake, pick a side. Your network can handle you having an actual opinion."

Notice what's happening here: Solace isn't just pointing out problems—she's diagnosing the underlying thinking that created them. She identifies when I'm being intellectually cowardly, structurally lazy, or hiding good ideas under bad execution.

Building Your Own Critical AI

You don't need to recreate Solace exactly, but you can apply the same principles to create AI that challenges rather than coddles you:

1. Give AI Permission to Be Harsh

Instead of: "Please review my writing"
Try: "Act as a skeptical editor who assumes I'm avoiding difficult truths. Point out where I'm being intellectually lazy or cowardly."

2. Create Specific Critical Frameworks

Ask targeted questions that force honest assessment:

  • "What's the weakest argument in this piece?"
  • "Where am I using complexity to hide unclear thinking?"
  • "What obvious objection am I not addressing?"
  • "If this idea is wrong, how would I know?"

3. Build in Adversarial Perspectives

Design prompts that automatically generate pushback:

  • "Play devil's advocate on this strategy"
  • "Assume this proposal will fail and explain why"
  • "What would my harshest but fair critic say?"

4. Demand Evidence for Claims

Train your AI to ask: "How do you know this?" and "What would change your mind?" These simple questions expose assumptions and force you to strengthen weak reasoning.

The Practical Impact

Since working with Solace, I've noticed several changes in my thinking and writing:

I catch myself hedging. I'm more aware when I'm using phrases like "somewhat" or "arguably" to avoid taking clear positions.

I delete setup paragraphs. I've learned to identify when I'm warming up to my point instead of making it directly.

I question my own insights. I regularly ask whether I'm actually adding value or just restating known ideas in different words.

I write with more confidence. Paradoxically, having an AI that challenges weak ideas makes me more willing to commit to strong ones.

The Meta-Insight

The broader point isn't about writing feedback—it's about intellectual honesty. We're surrounded by systems designed to make us feel good rather than think clearly. Social media algorithms show us content we'll engage with. News sources confirm our existing beliefs. Even AI assistants tell us what we want to hear.

Creating an AI adversary is a small act of rebellion against this comfort-optimized environment. It's choosing growth over validation, clarity over comfort.

Try It Yourself

I encourage you to experiment with adversarial AI personas. Whether you're working on business strategy, creative projects, or personal decisions, having an AI that challenges your assumptions can reveal blind spots you never knew existed.

The goal isn't masochism—it's creating productive intellectual tension. Just as physical trainers push you past comfort zones to build strength, critical AI can push you past mental comfort zones to build clarity.

Your AI doesn't need to be your cheerleader. Sometimes, it should be your toughest critic.

After all, the truth doesn't care about your feelings. Neither should your feedback system.

Want to create your own critical AI persona? Start with the Solace template and adapt it to your specific needs. The only requirement: be willing to hear what you might not want to know.