I’ve been sitting with a question for a few weeks that I can’t make go away by being busy. Which is inconvenient, because being busy is what I’m best at.

The question came out of a set of sticky notes I wrote after a long week — the kind you write at speed because the thinking won’t sit still, and then you stick them on the wall behind your desk and hope they’ll organise themselves. They didn’t. But they kept asking me the same three things, in slightly different order, every time I looked up from whatever I was shipping.

Is AI going to change who will be successful? Is it OK to filter people out? Do we need to bring people on the journey?

And underneath all three, scrawled in a different pen — the one I think I wrote last, after the others were already on the wall:

Am I feeling this because I’m in front?

The view from in front

A few months ago I wrote a piece called “The Friction Was the Thinking” — about how AI’s frictionless production strips out the judgment that used to live inside the effort. I identified three relationships people have with the same pressure: FOMO, brain fry, and production addiction. And I argued that the skill now is knowing which things earn the human — which work deserves the friction, and which can go without you.

I stand by that. But writing it surfaced something I didn’t fully reckon with in that piece: I’m writing from inside temperament three. I’m the person who shipped three apps on a weekend and then asked whether I should have. I’m the person who built a critic agent to prevent me from shipping too fast — and then nearly automated the critic. I enjoy this. The rebuilding, the constant novelty, the tool that was right yesterday being wrong today. That doesn’t feel like chaos to me. It feels like energy.

Which is why the question bothers me. Because when I say “the skill is now judgment about where friction belongs,” I’m describing something that comes naturally to a person like me — someone who finds the constant rebuilding exhilarating rather than exhausting. And I’m starting to wonder whether I’ve mistaken my temperament for a universal standard.

The new bar

Here’s what I keep seeing in the people who are succeeding with AI at speed, across teams and across companies: they can throw it all away and start again. The tool you had yesterday is completely different to the tool and skills you need today. The workflow you built last month is already obsolete. The mastery you developed over a specific prompt pattern or agent configuration is worth less than the willingness to abandon it and learn the next thing.

This isn’t a skill in the traditional sense. Skills accumulate. This is more like a disposition — the capacity to be perpetually a beginner without it costing you your confidence. To treat expertise as disposable and be happy about that. To find the restart exciting rather than demoralising.

I watch some people do this effortlessly. And I watch others — equally intelligent, equally hardworking, often with deeper domain expertise — struggle not with the technology but with the pace of obsolescence. It’s not that they can’t learn the new tool. It’s that they haven’t finished grieving the last one.

And I think that distinction — between people who metabolise constant change easily and people who don’t — is becoming a de facto filter in a lot of organisations, even though nobody has named it as one.

Is it OK to filter people out?

This is the question I keep circling. Because if adaptability at this specific speed is now the bar, then we’re filtering people not by competence, not by effort, not even by intelligence — but by psychological flexibility. By how much novelty-induced uncertainty they can tolerate before it becomes paralysing.

And the sharp edge of that: the people being filtered out often don’t know what they did wrong. It’s not like missing a deadline or failing a test. It’s more like the ground shifting under you while everyone else seems to be surfing. The slop ships, the review decays, the output looks polished but the judgment isn’t there — and from the outside, it looks like performance. From the inside, it feels like drowning in a room where everyone else is breathing fine.

The traditional answer in high-performance organisations is to hold the line on standards — to expect people to keep up. But that assumes the standard is visible. If the failure mode is invisible — if AI-generated slop looks indistinguishable from careful work — then people get filtered by an assessment they can’t see and don’t understand. That’s not holding a standard. That’s moving the bar without telling anyone.

Do we need to bring people on the journey?

If you’re in front — if you can see the production addiction, the review decay, the way frictionless output makes everything feel low-stakes — then you have a choice about what you do with that vantage point.

Option one: let it run. Natural selection. The people who adapt will adapt; the people who don’t will reveal themselves through their outputs. It’s efficient, it’s clean, and it has the specific elegance of requiring you to do nothing.

Option two: build mechanisms that protect the people behind you. Design the onboarding differently. Create the review structures. Make the failure modes visible before they cost someone their reputation. Take what you learned by being early — the near-misses, the critic you nearly automated, the moment you caught yourself bingeing on output — and turn it into something that prevents the same failure in someone who’s six months behind you.

I know which one I believe in. But I want to be honest about why option one is tempting. It’s tempting because it flatters me. If adaptability is the filter, and I pass it easily, then option one says I earned my position. Option two says my position is partly luck of temperament — and that creates an obligation.

The “in front” problem

Here’s the thing I can’t fully resolve, and I’m not going to pretend I can:

I don’t know whether I’m seeing something real or projecting my own experience onto everyone else. Both are possible. When you’re early to a pattern, your sample size is you. You notice the thing because you’re living it — but that doesn’t mean your experience generalises.

Maybe the three temperaments I identified aren’t universal. Maybe they’re the way my brain categorises my colleagues under pressure. Maybe “production addiction” isn’t a phenomenon — it’s just what I call Tuesday.

And maybe the discomfort I feel — the nagging sense that we’re filtering people unfairly, that the pace is unsustainable, that someone should build the mechanisms to bring others along — is the specific anxiety of someone who’s gone far enough in to see the edges. The person at the front of the queue sees the cliff first. That doesn’t mean the cliff isn’t real. But it does mean the cliff might look different, or be in a different place, than what they’re describing to the people behind them.

So I hold both. The observation is real — I’ve watched it happen, the invisible filter and the good people quietly washed out, never quite knowing what they did wrong. What I can’t be certain of is whether it generalises the way I think it does, or whether the shape I’m drawing is the shape everyone else would draw. The person who enjoys the chaos arguing that the chaos needs managing is, conveniently, putting himself at the centre of the solution. I notice that about my own argument. I just haven’t found the part of it that’s wrong.

What I’m left with

I don’t have a framework to offer here. That feels like a failure by the standards of the cultures I most respect — the ones that value mechanisms over intentions, where good intentions don’t work and mechanisms do. I believe in that. I just think this is a case where the mechanism hasn’t been designed yet, because we haven’t agreed on what we’re optimising for.

Are we optimising for the people who can surf the constant rebuilding? Or are we optimising for the broadest possible set of people to contribute meaningfully to work that’s changing underneath them every month?

Those aren’t the same objective. And depending on which one you pick, the mechanism looks completely different.

If you pick the first, you build harder, faster, more adaptive — and you accept that some people won’t make the transition.

If you pick the second, you build deliberately slower in places — you reintroduce friction where it protects people, you make the failure modes visible, you design for the six-months-behind person instead of the six-months-ahead person.

Here’s the part I keep snagging on: I’m not sure most of us have actually decided which of those two we’re optimising for. I think we say the second and measure the first — and I don’t think that’s bad faith, I think it’s that nobody’s made the choice out loud. Including me. And the people in front — the ones who can feel the gap between what’s said and what’s rewarded — will keep asking themselves whether they’re seeing clearly or just seeing from a particular vantage point.

I don’t know. I’m in front. I think the view is real. But I’ve been wrong before about the difference between my preferences and everyone else’s needs.

The sticky notes are still on the wall. They haven’t organised themselves yet.

This is a sequel to “The Friction Was the Thinking”. The question of who gets filtered out, and what we owe the people coming up behind us, is one I first started circling in “AI and Careers: What I Told My Daughters”.