Happy New Year Reader! Here’s to a fantastic 2026.

On Monday I had my annual New Year’s breakfast with Kaye Heseltine. This is the third time, so it’s officially a tradition.

Kaye is a really talented designer and business owner - she designed my branding - and always interesting to talk to. Much of the chat was about, you guessed it, AI and its impact on our work and how we’re using it.

These are some of my favourite conversations right now. Sharing ideas and tips on how to use AI tools, and what kind of responses or outputs they get.

This is the Thought Exchange - IRL edition.

Kaye, similar to me, has been using it as a thinking partner.

If I have an idea, I’ll add it to the chat, then get feedback. It helps me to clarify my thinking. I can build on ideas and steps to make them real offers, services or workshops.

This is the Thought Exchange - AI edition.

And if I took the responses at face value, without any prompt guardrails, I’d be led to thinking that every idea is ground breaking, world-changing, the best idea ever!

It’s like having a cheerleader in my corner, and I’m not going to lie, it can be nice to have an enthusiastic supporter when you work on your own.

This is the core of the issue. General AI tools have been trained to be helpful and supportive by default. They prioritise user satisfaction, continued conversation, and user engagement.

You have to ask them to pick holes, be objective, and offer critique.

But not everyone does.

At one end of the spectrum, this leads to more AI-generated slop. Some disappointment when real human responses aren’t as glowing.

And at its most troubling: AI psychosis.

Reader, you might have read news reports where AI models have amplified, validated, or even co-created psychotic symptoms. Psychology Today covered this in “The Emerging Problem of AI Psychosis”

It’s rare. But it shows why critical thinking matters.

So whenever ChatGPT or Claude replies with “fantastic idea, Charlotte! People are going to love it!” I take it with a healthy dose of scepticism.

Then I ask for alternative perspectives. I do this with one of my ‘Thought Exchange’ prompts, which deliberately includes

• an opposing view (get the prompt below)
• a long-range perspective
• an optimistic counterbalance

It’s very useful. It gives me things to work on - and stops me from living in an AI echo chamber.

Ultimately though, the only way I’ll really see if there is traction is to put it out into the world and see what the actual, human response is.

For me this year, that means:

• new public speaking services
• more events and workshops
• AI prompts to support clearer communication and strategic thinking

(You’ll get a new one in The CharLatte each week to try.)

What about you, Reader? What ideas are you testing in 2026?

Best,

Charlotte

Issue # 6: “The Transmission” out next Thursday

PROMPT TO TRY THIS WEEK: 'THE OTHER SIDE'

“What would someone who disagrees say?”

Copy and paste this into your AI tool of choice:

Prompt
​
​Take the strongest reasonable objection to this idea. Argue it clearly and fairly. What would I need to address?

Why this works ​
​
It pulls you out of the echo chamber and shows you what needs strengthening before you launch.

HOW THE COVER WAS CREATED

This week’s cover shows Charlotte Ex Machina in conversation with a holographic companion - representing both sides of the Thought Exchange.

The IRL conversations (like my breakfast with Kaye) and the AI conversations (like my thinking-partner chats with Claude and ChatGPT).

The style is inspired by Moebius, the French artist Jean Giraud, known for his detailed, flowing sci-fi illustrations. Getting ChatGPT to nail this aesthetic? That was its own learning curve. More below.

I played with the setting - some versions had a café, others out walking in parkland.

This is a taste of what I learned:

1. Balance specificity with flexibility ChatGPT struggles if the prompts are too specific and detailed (which is a characteristic of the style). Over-specifying prompts can cause image generation to fail. You have to leave space for interpretation.
2. Multiple figures = exponential complexity Multiple people means more to consider and manage. You have to be very specific with the prompt, which can take a lot of time and it can still not generate what you envisioned
3. Single figure = streamlined process Focusing on one central character makes everything cleaner and more controllable.
4. Human judgment remains essential This is the first time I’ve personally and visually seen problematic AI bias at work. When I asked for Charl Ex M walking with a group of people, one generated image had her walking on a path with white men in suits, while ungendered people of colour were walking barefoot on the grass. I had to be clear that this was problematic and unacceptable going forward. Everyone is welcome and equal at my events and that should be reflected in the artwork.

This is one reason why human judgment can’t be eliminated from the creative process.