Co-authored by Harry Sharman 🧠 & Josh Madej Published in Beautiful Thinking)
The Conversation We Can’t Stop Having
If AI stopped improving today — no smarter models, no breakthroughs, no AGI — would you still have a job five years from now?
That’s the question that keeps dragging us back into the same weekly conversation.
It started as curiosity — two people close enough to AI to see what’s under the hood. Then it became modelling. Then it became a slightly masochistic ritual in which we calmly walk through what increasingly looks like the expiry date of our own careers.
We sit there, feeling the air leave our lungs in slow motion, as if we’re politely taking turns delivering a gut punch and then analysing the bruise. And occasionally we ask the obvious question: is this foresight… or have we simply developed a taste for the impact?
To be clear, our conversations absolutely wander into Terminators, AGI and runaway superintelligence. But one of our wives keeps reminding us — quite reasonably — that none of those are certainties.
So here’s the thought experiment.
Freeze AI exactly where it is today.
No godlike leaps. No artificial super-scientists. No “country of PhDs.” Just the systems we already have — which, incidentally, would be unprecedented in technological history because no transformative technology has ever politely stopped improving.
Right now, AI is pretty good. Occasionally wrong in oddly random ways. But capable.
It can already write code. It can control your screen. It can click, scroll, type, speak and listen. Every way you interact with a machine — keyboard, mouse, voice — it can interact the same way. Often faster. Often tirelessly. Often at scale.
Most white-collar work happens on a computer. AI operates on computers.
That overlap isn’t philosophical. It’s mechanical.
Even if models never became more intelligent than they are today, the remaining gap isn’t IQ. It’s orchestration — wiring agents together so that what a human does across five tabs and three meetings becomes a system that simply… does it.
And plumbing gets built.
If your job consists largely of moving information between digital containers — documents, decks, emails, spreadsheets, briefs, analyses — with some thinking in the middle, there is a plausible shelf life attached to it.
Not tomorrow. Not next quarter. But within five years? Even with AI frozen at today’s level, we struggle to see how large swathes of that work remain economically rational to pay humans to do.
Institutional inertia will slow it. HR moves slowly. Managers like managing humans. Culture doesn’t evaporate overnight.
But slow is not safe.
This doesn’t feel like disruption.
It feels like countdown.
The Economic Arithmetic
There’s a word people use.
Displacement.
It sounds calm. Measured. MBA.
What we’re circling around is a question we don’t hear asked clearly enough: if large parts of knowledge work become automatable, where exactly are we all getting displaced to?
Because at some point knowledge-based companies face a decision.
Hire a human for $100,000 a year who takes 1,000 times longer.
Or hire an agent — or agent swarm — for $1,000 a year. Or $1. That produces 90% of the work. Sometimes better. Sometimes imperfect. But fast. Scalable. Tireless.
In competitive markets, no one chooses the slower, more expensive option out of nostalgia.
AI doesn’t need to be flawless. Humans aren’t.
It just needs to be good enough.
And if it delivers 90% of the output at 1% of the cost, the economic gravity is obvious.
We’re not talking about distant superintelligence. We’re talking about systems that exist now, improving incrementally, being wired together by engineers who are very good at plumbing.
That’s what keeps us talking. And occasionally staring at each other in silence.
The Part Nobody Has Solved
What happens when a significant proportion of knowledge workers can be replaced by agents — and the economy continues to grow?
Historically, when jobs shift or disappear at scale, societies wobble.
But this time productivity may rise because displacement has occurred.
That’s new.
And it’s being steered — intentionally or not — by a small number of frontier AI companies building systems that improve every quarter.
No conspiracy. Just concentration of capability.
Once you see that, you can’t unsee it.
And once you can’t unsee it, a harder question appears.
What do you do with knowledge like this?
The Genetic Test
Years ago, one of us chose to take a genetic test. The reasoning was simple. If there were a high probability of early-onset Alzheimer’s, that would change how life should be lived.
You wouldn’t optimise for retirement.
You’d optimise for time.
More time with your spouse. With your kids. With friends. Different financial choices. Different priorities.
When this was mentioned at work, the response was almost unanimous:
“I’d rather not know.”
Knowing attaches a clock. It turns a horizon into a countdown.
And telling people about AI displacement does something similar.
It attaches a clock.
The Mormon Paradox
Here’s the theological detour that clarifies the dilemma.
In Mormon belief — simplified, but accurate enough — if you hear the truth of the religion and reject it, there are eternal consequences. You knew. You chose otherwise. Down the fiery steps you go.
But if you never heard it — because you grew up somewhere missionaries rarely reach — then when you die you are given a choice.
At the gates.
Mormonism this way — eternal paradise.
Not-Mormonism — those flaming steps.
In that world, who isn’t choosing the gates?
Exactly.
Ignorance becomes strategically advantageous.
Now apply that to AI.
If someone never hears about large-scale job loss, they live peacefully. They plan promotions. They build pensions. They assume continuity.
If we knock on the door and say, “There’s a meaningful probability your role won’t exist in five years,” they cannot unhear it.
They carry the clock.
And here’s where the analogy breaks.
In the Mormon story, there’s a choice at the end.
In the AI story, there may not be.
There’s no pearly gate moment where you opt back into economic relevance.
So should we still be pressing the doorbell?
Should we be standing outside in crisp white shirts and skinny black ties — yes, hear the opening of The Book of Mormon in your head, “Hello!” — clutching our AI prophecy black book and asking if we can tell you the most important thing you’ll ever hear?
Because we’re not offering salvation.
We’re offering awareness.
And awareness might not change the ending.
The Anti‑Thought Leader Problem
We write about AI. We speak about it. And increasingly we ask ourselves whether shining light on trajectory without offering control is a public service or mild cruelty.
When people say, “What do you want me to do about it?” the honest answer is thin.
Reskill — into what?
Pivot — into which defensible domain?
Trust policy — from whom?
We have probabilities. We have exposure. We have pattern recognition.
What we don’t have is a clean solution.
Which raises the deeper question: is knowing good if it reduces comfort but doesn’t meaningfully increase agency?
What Knowing Is For
The reason to take the genetic test wasn’t control over the outcome. It was control over how to live before it arrived.
Perhaps AI functions similarly.
Maybe the value of knowing isn’t prevention but choice. If you believe the runway is shorter than assumed, you might optimise differently — for time instead of status, relationships instead of accumulation, optionality instead of linear promotion.
Or perhaps that’s just a comforting story we tell ourselves while nursing sore metaphors.
We genuinely don’t know.
The Support Group
So for now, we’ll continue our weekly ritual. We’ll compare notes. We’ll model timelines. We’ll sit there, feeling that familiar gut punch again, calmly examining it and wondering whether this is foresight… or just refined masochism.
If you’ve been having similar thoughts quietly, you’re not alone.
We’re considering starting a support group.
The AI Apocalypse Support Group.
Or perhaps more honestly:
The Gut‑Punch Support Group — impact tolerant members welcome.
(Probably) No cult. No silver bullets. No ten‑step framework.
Just honest conversation about whether knowing is a gift, a burden, or both.
If you’d join that conversation, let us know.
We promise only manageable bruising.