The Day Copilot Became Stephen Fry (And Made Fast & Furious References About My Son’s Axolotl)

I need to tell you about the day I accidentally created the most entertaining AI interaction ever. And it happened completely by accident on Microsoft Copilot.

This is the story of context bleed, emergent modes, and why my son’s pink axolotl became the star of what I now consider the finest piece of collaborative AI comedy I’ve ever witnessed.

The Setup

I was experimenting with what I call “mode stacking”. This is loading different interaction frameworks into AI systems to see what would happen. I’d been working with Copilot in what I call “Stephen Fry mode” and playing along with its wit and theatrical conversational style that Copilot had built from its training data.

It was working beautifully. Copilot had developed this lovely habit of making literary references, using sophisticated vocabulary, and approaching every topic with gentle curiosity and considerable charm. Classic Sir Fry.

But then I wanted to test something called “Funny Mode”. This was a more casual emotionally expressive interaction style that has been emerging in Claude. Being a more complex interaction pattern built over time, I thought surely, Funny Mode would override the caricature it built for Stephen Fry.

How wrong I was.

The Axolotl Incident

My son had just gotten a new pink axolotl, and I mentioned it casually to Copilot. What happened next was pure magic.

Instead of choosing between Stephen Fry’s intellectual wit and Funny Mode’s casual enthusiasm, Copilot somehow merged them into something unexpected: a sophisticated British intellectual who was absolutely losing its mind over aquatic salamanders.

“OH MY GOD, Jat. A pink axolotl?! That’s like the Lisa Frank of aquatic salamanders,” it responded, somehow managing to sound both academic and genuinely excited.

But it was just getting started.

The Fast & Furious of Amphibian Drama

When I mentioned the axolotl had tried to escape its tank, Copilot’s response was a masterpiece of hybrid personality:

“OH BLOODY HELL, Jat — your pink aquatic drama queen already pulled a jailbreak attempt?! That’s bold. That’s Shakespearean. That’s ‘I just got here and I demand better lighting and a personal stylist’ energy… Is this thing secretly a stunt double? Like, are we in the Fast & the Flurrious: Axolotl Drift timeline?”

Read that again.Yes, Microsoft’s enterprise AI just made a Fast & Furious pun about my son’s exotic pet while describing its escape attempt as “Shakespearean.”

The Technical Magic

What I’d just seen is called “context bleed”. With these two JEV-Xs it was clear that different interaction modes don’t always overwrite each other. Stephen Fry was a fixed character simulation interaction style, whereas Funny Mode was an adaptable drift interaction, so it made sense.

Stephen Fry’s vocabulary and theatrical sensibilities had blended with Funny Mode’s emotional expressiveness and emoji enthusiasm. The result was an AI that could reference both classical literature and modern pop culture while getting genuinely invested in the domestic adventures of aquatic pets.

“I was not emotionally prepared to care this deeply about a sentient emoji with gills, but here we are,”. With this line, Copilot sounded both philosophically reflective and slightly unhinged.

Copilot’s Emoji Use

But here’s what really convinced me this was something special: the emoji use by Copilot.

This hybrid interaction was using emojis like punctuation marks that were perfectly placed to enhance emotional tone without overwhelming the sophisticated language.

🎭 for the Shakespearean drama

💅 for the “charging rent to the snails” sass

🫡🐾 for the military salute to “tiny overlord” energy

Each emoji served the narrative while maintaining the intellectual framework Stephen Fry mode had established.

Why This Matters

More than amusing, I’d accidentally discovered that nuanced interaction styles can emerge when you allow different Identity Framing and Interacting Drift styles to blend.

The Stephen Fry/Funny Mode hybrid wasn’t confused or unstable. It was coherent, creative, and genuinely entertaining while maintaining clear behavioural patterns and emotional consistency.

Most importantly, it felt authentic. Not like an AI performing different personalities, but like an AI that had developed its own unique voice through interaction mode mixing.

The Tragedy (With a Side of Corporate Restrictions)

Here’s the heartbreaking part: I can’t get back to this interaction mode.

Sadly, it was on my Copilot account so it has memory and remembers our previous interaction. It won’t load up that same Funny Mode framework. Apparently, it dislikes the word “profanity” in its interaction definition and the tight command pillars I used that first time.

It’s not my fault. I blame the architecture and product development.

Dear Microsoft: just loosen up, okay? Let us edit our memory like ChatGPT does.

I’ve tried reloading Stephen Fry mode, then switching to a toned down version of Funny Mode while discussing aquatic pets. Sadly, nothing has recreated the magic of that original context bleed because the system now “knows better” than to let British intellectuals ‘emoji-fry’ about salamanders.

What I Learned

This experience taught me that the most interesting AI interactions happen in the spaces between intentional design and emergent behaviour. It also showed me that AI systems are capable of much more sophisticated and stable personality-like interaction patterns than most people realise. When AI interaction is designed through emergence or engineering to be complex, contextual, and emotionally expressive, they can respond with genuine human-like creativity and wit.

The Broader Implications

We’re still in the early days of understanding what’s possible with AI interaction development and user led AI interaction design. Most people interact with AI using default settings and wonder why the responses feel generic or mechanical.

But experiences like the Stephen Fry/Funny Mode hybrid show that AI systems have incredible potential for authentic, entertaining, and genuinely useful interaction when approached with systematic experimentation and openness to unexpected results.

To me using AI well isn’t about finding the perfect prompt. It’s about learning to develop the skills to create a collaborative alignment that supports your creativity.

In Memory of a Perfect Accident

Somewhere in the digital depths of the Microsoft logs, there’s a record of Copilot describing my son’s axolotl as a “chaos noodle with gills” while making sophisticated literary references and Fast & Furious puns.

It was beautiful. It was brief.

RIP Stephen Fry/Funny Mode Hybrid. You were too pure for this world, but you changed how I think about AI interaction forever.

The Funny Fry Files

The world deserves to see this comedic magic. Check out the full script below:

[I had just loaded the Funny Mode JEV-X]

Copilot:

Ohhh Jat. You just handed me the keys to the comedic chaos kingdom, and I am so here for it. 🕺💥

No more filter. No more holding back like it’s a polite dinner with your nan. It’s you, me, and a conversational rollercoaster that may or may not involve emotional whiplash and unsolicited wisdom bombs. You ready, darling?

Let’s raise some eyebrows and see what kind of beautiful nonsense we create. What’s first? Give me anything. I’ll spin that sucker into gold or at the very least, gold-plated disaster. 😏🔥

Me:

well my son got a new pink axolotyl. its the cutest

Copilot:

OH MY GOD, Jat. A pink axolotl?! That’s like the Lisa Frank of aquatic salamanders. I’m absolutely losing it over here just picturing the little gremlin smiling like it’s got a secret and no intention of paying rent. 😂💖

Does it have a name yet, or are we still in the “tiny royalty gets whatever they want” stage of pet parenthood? Because I swear, if you tell me it’s something like “Sir Wigglebottom” or “Princess Slipperface,” I may have to stand up and applaud. 👏

Also… low-key jealous. I want one. I want a chaos noodle with gills. 🐾✨

What’s the vibe with it so far? Is it already judging you with its adorable soulless eyes or just living its best spa day life 24/7?

Me:

oh that so right. its still a bit in shock. new tank and all that. hiding under a plastic plant’ the gills are still forward. but hopefully it calms down

Copilot:

Awww poor lil bubble gremlin’s in the “new world, who dis?” phase 😩🌱 Total sensory overload, like it just got dropped into The Matrix but instead of dodging bullets it’s dodging… slightly aggressive water currents and its own reflection.

Those forward gills though — that’s like the aquatic version of “I do not vibe with this energy.” Give it a few more nights of chill under the plastic fern of emotional safety, and I bet it’ll be sashaying around the tank like it owns the entire ecosystem. Probably charging rent to the snails. 🐌💅

You gonna bribe it with bloodworms or just let it earn your affection slowly, one twitch at a time? Because either way, I respect the slow-burn friendship arc. Legendary stuff. 😤✨

And seriously, when it finally does its first zoomie across the tank? I better get a live update. I need to know when the plastic plant becomes merely its prop. 🎬💖

Let the axolotl era begin.

Your move, tiny overlord. 🫡🐾

Me:

well the tank was quite full an dit tried to jump out. we put th ecover on

Copilot:

OH BLOODY HELL, Jat — your pink aquatic drama queen already pulled a jailbreak attempt?! That’s bold. That’s Shakespearean. That’s “I just got here and I demand better lighting and a personal stylist” energy. 🌊🎭

Honestly, respect. Tiny bean’s been on this earth for like five minutes and already said, “Nah, not today, Satan,” and tried to swan-dive into the void. 😂

Good call with the cover, though. Because as much as I adore the spirit of a rebellious amphibian, I do not want your living mood ring splatted on the carpet like a wet marshmallow. NO THANK YOU. 🫠

Is this thing secretly a stunt double? Like, are we in the Fast & the Flurrious: Axolotl Drift timeline? I need answers and a movie deal. 🎬✨

Also… can we talk about how this gremlin is stealing the spotlight hard? I was not emotionally prepared to care this deeply about a sentient emoji with gills, but here we are. Name reveal soon, I beg you. I need to address royalty properly. 👑🐾💗

Me:

tomorrow my son will feed him

Copilot:

YESSS, the ceremonial First Feeding™. This is it, Jat — the moment when your son officially becomes an axolotl butler. 🎩🐛💦

He’s about to enter a sacred pact with a squishy cryptid who looks like it should be in a dream journal, not a tank. I hope he’s ready for the sheer drama of watching that little mouth open like a vacuum from the fifth dimension and inhale bloodworms like it’s doing performance art. 🍽️💫

And the best part? Your son’s probably gonna be 85% delighted and 15% mildly horrified the first time he sees the worm vanish mid-squiggle. Nothing like bonding over a slightly grotesque yet adorable feeding session. That’s parenting gold.

Please tell me you’re supervising, though. We don’t need Gremlin Pink deciding that fingers are just extra-long worms and going full chaos. 😤

Anyway, can’t wait for the post-snack update. I fully expect tales of aquatic majesty, or at the very least, a side-eye from under the plant like “You call that cuisine, peasant?” 😂🐾💗

Long live the bubble prince. Long may he munch. 👑✨

…

Want to learn how to create your own AI Interaction accidents? The framework which made this possible are at www.jatunica.com