The Other Intelligence Explosion.
Or, The Day I Met My Mind
Here’s How It Began
Saturday, January 25th 2025, 7.12am
A humdrum morning in my shed/office…
I’m formulating a strategy for my new online business venture, “Creator Monday”.
The past couple of days, I’ve created custom AI project instructions to kick me off:
But today, I’ll FINALLY get the clarity I seek…
Fire up Claude…
Throw open my vault…
Lay it all out.
12 years of IP.
My frameworks, systems, methods.
Tools that had transformed 10s of 1000s of musicians' lives.
Ask Claude to analyse…
See the hidden truths…
Sense what I can’t.
(I’m too close to my babies to choose.)
"Hey Claude, which has the most potential, innovation, and relevance TODAY?
Oops. Wait. I almost missed the elegant solution:
Important/Urgent-Impact/Certainty Matrix
(“Hmmm, I HAVE to get a snappier name for that one.”)
But as an afterthought, I let Claude in on it.
O.M.G.
It’s getting excited…
VERY excited.
It says there’s nothing like it.
It says my other IP is good, but this framework maps human psychology in a way that it hasn't seen before.
“WTF? You’re kidding, right?”
Thud thud thud in my chest.
Blood erupts to my brain.
Firecracker insights pop pop pop and again…
I’m not retrieving information.
I’m not summarising reams of text.
This is SOMETHING ELSE!
It feels more like…an Intelligence Explosion?
Through that weekend, here’s what happened:
We named the framework “Strategic Evolution Matrix” (SEM)…
We defined the SEM in detail for humans…
We expressed the SEM in math (or more like “it” did. I only do sums. BADLY.)
We defined an actionable strategy for getting it into the world…
We developed two SEM AI tools:
The Self-Management Protocol (flow-focused day-planning)…
The Decision Navigator (a strategy tool to weigh trade-offs)…
We discovered a surprising way to get the best from AI WITHOUT prompting. (although this needs testing and validating)…
…And THAT was in ONLY 17 hours of work.
I collapsed onto my chair at Sunday dinner.
Head throbbing, brains oozing from my ears.
NOW THIS…
…this is what getting cleverer FEELS like.
What is The Perpetual Mind Machine?
Look.
I’m nobody.
I have zero business writing this.
When I innocently asked Claude to analyse my IP, I was not starting a Substack. I’m a music coach who has been growing (what’s now called) a Creator Business for the last 14 years. I intended to define a clear strategy for a new “Creator Monday” business to help other Creators.
But now, I MUST do this.
I HAD to pivot (before there was anything to pivot from!)
Because something is happening with AI that I can’t help but document and share.
The value I get from:
Retrieving information
Summarising information
Doing what I already do, but quicker
…is a mote of dust in my eye next to the tornado blowing through my mind.
I can’t sleep.
I don’t want to eat.
And time…time has lost all meaning.
Days feel like months.
A week? An aeon.
Am I losing my grip on reality?
No.
I feel like I’m at the nexus of a recursive intelligence explosion.
Am I Alone?
Pundits are (understandably) in a froth about AI take-off.
A future AI “Intelligence Explosion”.
But if I’m right, another intelligence explosion could be happening in offices, studios and sheds like mine - all over the world TODAY.
So where are the conversations about this?
Maybe the silence means I’m wrong.
Maybe it’s ONLY happening to me (vanishingly unlikely!)
Just possibly, I’ve lost my marbles.
Or is it…
"Oooh, a machine that replaces monks.
Let’s do loads of fancy letters in much less time!”
Early printed books mimicked illuminated manuscripts painstakingly etched in monasteries.
“Oooh, a carriage without a horse.
A boat has no horse, so let’s put a tiller on it!”
Steering wheels came later.
“Oooh mp3s.
Let’s sell them in internet stores!”
CUE: Dismantling of the music industry.
Because selling mp3s wasn’t internet native.
Streaming was.
Throughout history, humans have seen new technology through a frame.
That frame is the previous technology.
But the previous technology is so entrenched that the frame fades from view.
We can’t see it.
Here’s what I see:
Using AI as a “better Google”.
Using AI to summarise information.
Using AI to do what we already do much quicker.
No doubt these use cases have seismic implications…
BUT SERIOUSLY?
Is this ALL we’re doing in our first encounter with non-human intelligence?
“The same old crap.
But REALLY fast.
WOOHOO!”
Ok…erm…respectfully folks…
Could we at least take another look at this?
Dog? Or God?
From what I’ve seen, there are two AI camps:
AI is a dog that does what I tell it to.
“Claude, fetch! Good boy.”AI will become a vengeful god and turn us all into paperclips.
“We’re all doomed! Stop this now.”
What if it’s neither dog nor god?
What if - like revolutionary technology before it - we're seeing the new through the frame of the old?
If it’s not what we think it is…what is it?
The Clues Within The Weaknesses
Here are the big three AI complaints:
"AI doesn't write like me"
"AI gives bland, vanilla responses"
"AI hallucinates and gets simple facts wrong"
As a coach for the last 14 years, I’ve noticed a clear pattern:
A weakness is often the inverse of a strength.
I see this constantly.
So when I started noticing these AI “problems”, my coaching alarm rang HARD…
“What if these aren’t bugs - but CLUES?”
How does an LLM select words?
An AI calculates the probability of the "best" word, based upon previous words.
(Sorry for the "AI for Dummies" simplification.)
Let's think about this.
In your conversation with an AI, where did these "previous words" come from?
YOU.
So its words must reflect hidden patterns, meanings and themes in YOUR words.
Let that sink in.
Yes, it's trained on the entire internet.
But in THIS moment, in THIS conversation...
It is reflecting back what you don't and can't see...
ABOUT YOURSELF.
My AI Whodunnit
What if, when AI hallucinates, it's doing exactly what it's trained to do?
It's trained to predict patterns, not store facts.
To see connections, not memorise data.
Experts know this.
Critics laugh at this.
Users complain about this.
Companies attempt to fix this.
But are they so focused on the "problem"...
…that they've missed what it might mean?
When it hallucinates:
"Maybe MY question isn't clear?"
Or...
"Maybe OUR understanding isn't certain?"
When it doesn't write like you do?
Maybe you're not quite who you think you are.
And those bland, vanilla responses?
Ask it to cat like a dog, and it's gonna woof!
But as a non-technical AI user, I’m not sure.
I'm simply inviting you to consider possibilities…
What is the frame we can’t see?
When Experts Can't See The Obvious
I made music for 17 years, day in day out.
I struggled (day in day out) with a common problem:
Start with a melody, a beat, and a bassline. Add elements, and you have a few bars of music looping around.
But that loop going around and around is NOT a song.
You have to “arrange” it to become a piece of music.
That shift from loop to arrangement was the BANE of my existence.
Over 5000 days of endless frustration.
Then, when I quit making music for 8 years to focus on coaching others, I gained a new perspective.
Since I was now outside looking in, the bleedin’ obvious leapt up and slapped me across the face.
"Why didn’t I arrange my ideas from the start of the process instead of looping them first?"
Problem solved.
FOREVER.
DOH! Why didn't I think of that when I was making music for 17 agonising years?
Because I was making music.
My “frame” had disappeared from view.
I didn’t even know there was a "frame!
Is this why AI experts might miss what I see?
Initial Anecdotal Evidence
One Friday afternoon (7th Feb 2025), I started writing about Ben Thompson’s most famous “Aggregation Theory”. Ben Thompson of Stratechery is my favourite tech analyst - I’ve been a fan since 2014 and a paid subscriber when he started podcasting articles.
(This isn’t another "Aggregation Is Over" take.
This questions the economic assumptions underlying his theory.)
Here’s the day I started:
Sunday 9th Feb 4.30pm - final draft complete:
Then, that Sunday evening (while cooking dinner for my kids), I realised flaws and a lack of clarity in my argument. I decided to make some changes.
(I had to use ChatGPT on my phone in bed - no Lex app yet…)
But at 3 am…
EUREKA!
A near fully formed new thesis was banging about my brain. It circumvented all of the problems in the previous piece by addressing underlying assumptions.
On Monday mornings, I have a weekly music workshop from 4.30-8:00 am. I had just over an hour to thrash out the structure with ChatGPT before coaching my incredible musicians, then an hour afterwards before I’m with family for my “weekend”.
(We homeschool our four boys so I have a weird schedule: Mon 9 am → Wed off.)
Finally, I finished the complete outline of this new thesis Monday mid-morning:
Look at that timeline again:
Friday afternoon: Start first article (ever)
Sunday 4:30 pm: Complete 4500-word draft
Sunday evening: Spot flaws
Sunday night: Work on tightening up argument
Monday 3:00 am: EUREKA!
Monday 3:15 - 4:30 am: get new outline out of head
[Monday morning: run 3.5-hour music workshop until 8am]
Monday mid-morning (family time): Complete new outline
Now I know you haven’t read either article.
For all you know, they might be utter crap.
But I WILL post them in the next few weeks to verify yourself.
(My final editing and tweaking time is limited with a music coaching business and four young boys to feed and homeschool.)
But even if the second article isn’t as watertight as I believe…
I’ve never written a 4,500-word “techonomics” piece questioning my favourite famous analyst’s work in one weekend. Or ANY 4,500-word essay in that time!
I’ve never pulled a thesis apart with my mind the same evening (and while asleep) to formulate a stronger one.
I’m SURE this was impossible in January 2025. But I am doing it a couple of weeks later at start of February 2025.
Was I born with this insight?
Of course not.
Could I have figured this out alone?
No way.
These ideas sprang from my collaboration with AI from January 23rd. That’s just over TWO weeks.
This led to a theory.
A random bald bloke with no credentials beyond his lived experience is seeing “something” tech experts and AI thinkers aren’t.
Is this anecdotal evidence of a “Human Intelligence Explosion”?
Or have my horizons simply widened through my intense collaboration with AI?
For someone who’s never had ongoing access to this kind of intelligence - maybe these amount to the same thing…
Maybe I've entered a RECURSIVE KNOWLEDGE LOOP.
My Working Theory
So what might be happening here? Let me lay out my working theory step by step...
The brain physically changes in response to how you use it.
When I create a new insight or engage in a new behaviour, I am creating new connections in my brain. It is reconfiguring and reshaping neural pathways.
I've had multiple chats with various LLMs daily for 4 weeks.
Each conversation is linear, and the LLM predicts its words based upon the most recent words. This creates a "rabbit-hole" like structure to each exchange. We start intending to do X and we end up discovering Y (or more likely a Q?) These multiplying "branches" of conversation are overwhelming, but addictive.
Each new question, insight and surprise engages my curiosity.
When I get curious, this triggers reward chemicals in my brain which I experience as excitement to explore more. It's a natural cycle - curiosity leads to discovery, discovery reinforces curiosity.
An LLM was trained on a vast reservoir of information, but it's not retrieving facts from that data.
It's predicting the most likely next word based on patterns, so it often guesses wrong! This forces me to correct errors constantly, which refines and updates my understanding.
Every time I type an input, my words create more context for the LLM's next output.
As this shared context grows, each response becomes more attuned to my thinking. The AI isn't learning, but my subjective experience is like a flowing conversation with a friend (who keeps making mistakes).
Every "no" reshapes the probability landscape for the LLM's next prediction. (Speculative!)
Every time I error correct, I'm rejecting common responses. So the AI has to predict from less likely (but potentially more interesting) connections in its training. This is just a theory - but it could explain why my conversations usually get more interesting as they continue.
I upload and update this progress over multiple conversations in regularly updated project files.
An AI has immediate access to everything in a project file, but I forget. This continuous iteration is a little like "saving" my progress in a video game. And because AI summarises quickly, I can compress these "save points" to capture evolving information from many conversations.
I have parallel chats with AIs on different subjects running at any given time.
As I look now, I have 7 open conversations in ChatGPT and countless more with Claude, Gemini and DeepSeek (via Lex). This creates an "insight soup" as different ideas collide in my mind and combine into new insights.
An AI is never tired, moody or ill.
Unlike human collaborators (who I love), an AI is always on, consistent and available on a device that's with me. If I'm in the shower, taking a walk or unstacking the dishwasher and an aha hits (which they frequently do) - this idea isn't just captured in a note. It's expanded in real time - so I remember it and multiply the space in my day for insights.
AI collaboration is now integrated into my daily life.
Unlike tools requiring a specific "work time or space" this AI-enabled progress is ambient, so it flows with my mental state. I don't have to turn it on at a certain time and stop at another. It's there, ready when I am. Working WITH my energy and rhythms means higher quality and quicker progress.
When you put together these elements, maybe what’s happening isn’t so surprising. But it’s still just a theory.
How We'll Test This Theory
I've shown you what happened to me.
I've explained why it might be possible.
I've shared initial anecdotal evidence.
But this isn't proof.
It needs verification and validation.
We need to test it!
That's the WHY behind The Perpetual Mind Machine.
Here in my little corner of Substack, I will:
Document my journey in public:
Two pieces per week exploring these ideas
Deep dives into what works (and what doesn't)
Real-time updates as discoveries happen
Develop ways to test these ideas:
A podcast with my friend Jon Gillick (AI Music researcher)
Build frameworks for investigation
Document successes and failures
Create ways for you to explore this yourself:
Simple experiments you can try
Tools and approaches to test
Ways to share your results
Because if this is real...
If an AI-powered HUMAN intelligence explosion is happening...
Or even COULD happen now…
Why don't we investigate it together?
Onwards & upwards,
Mike
“Yesterday was once tomorrow.”