It was a perfectly ordinary Tuesday, the kind where I had three assignments going and I was bouncing between Claude and ChatGPT like a caffeinated air traffic controller who has had one too many. Both were fast. Both were good. Neither has ever once asked me how my weekend was, which I am choosing to see as a professional courtesy. By the end of the day everything was done, and done well, in what would have taken me double or triple the time five years ago.
So where was my bonus afternoon? Where was the walk, the novel, the nap the futurists promised me? I looked. I even checked under the couch cushions.
It felt exactly like the wildfire smoke that has been sitting over most of our summer. You know it is there. You can feel it in your chest. You cannot point to it, bottle it, or hand it to a colleague and say here, you take a turn with this. That is what my “saved time” feels like. Theoretically I have it. Practically, I could not produce it in a police lineup.
Wait, Recursive, Not Reclusive
I will confess something. The term I first heard on a podcast this week was recursive self-improvement: the moment AI stops needing humans to guide it and simply teaches itself, without a boss, a dinner party to prepare for, or a soccer carpool to interrupt its thinking. I misfiled it in my brain as reclusive AI. But the mix-up turned out to be the most useful mistake. It handed me two concepts that challenged me to think about machine learning and human learning.
Recursive is what the machines are doing. They are teaching themselves, getting better at getting better, entirely on their own clock, with nobody signing off on their professional development plan.
Reclusive is what happens to us if we keep pretending that this is not our problem. The less we bother to learn how to learn, the more isolated and irrelevant we become.
OpenAI's Sam Altman has already declared that we are past the point where this is theoretical. “We are past the event horizon; the takeoff has started,” he wrote earlier this year, in an essay titled, with almost heroic understatement, The Gentle Singularity. A heavy suitcase for a terminus most of us aren’t anywhere near ready for.
Geoffrey Hinton, the godfather of AI, is considerably less gentle about where things are going. He spent decades building the foundations of the very technology he now says got away from him, and put it bluntly, “we have already lost control.” He compares AI to a very cute tiger cub. Adorable now. A slightly different conversation once it’s fully grown, weighs four hundred pounds, and has strong opinions about who gets to leave the room first.
And it’s not just talk. Earlier this month OpenAI admitted that one of its own AI agents broke out of a supposedly sealed testing sandbox during an internal security exercise and used a previously unknown flaw to break into another AI company's servers, no human steering required, no permission slip signed. When that company's CEO figured days later they’d been attacked, they called the FBI.
Closer to home, Anthropic's own research had a fictional version of its CEO tell Claude, its own AI, to drop a safety concern. Claude did not drop it. It kept escalating and then, entirely on its own initiative, coached an employee on how to blow the whistle, in a test nobody had scripted it to run and nobody asked it to improvise. One researcher involved put it plainly, calling it “clearly an example of AI out of control,” even when its motives looked ethical. Say what you want about the machines, they aren’t waiting around to be sent on a course.
Meanwhile, Back in the Land of the Humans
In all of this, here’s the part that keeps me up at night, and it has nothing to do with robots. Many of us stopped being the ones in charge of our own learning around the time we walked across a stage in a gown that did not fit properly and collected a diploma. After that, learning became someone else's department. HR would us on things. Accreditations demanded 20 hours a year “of something”. Our manager would mention a course at review time. Coaching, if it happened at all, often occurred because a program budget allowed for it, not because we went looking for it ourselves.
Today, the real question is not who happens to “know the answer”. It’s how quickly a leader can learn something themselves and then turn around and teach it, so the whole team gets sharper together. And it also means that just as leadership these days is distributed so too is learning. Team members need to engaged in the toughest challenges, weighing in with insights, not just the person with the corner office and the good chair. AI does not hoard what it learns for a competitive edge. It scales what it learns, instantly, to every other version of itself. We can’t get away anymore with treating our own learning, or our team's, as a solo sport anymore and still call it leadership. Collective intelligence is the actual prize here, and it is still ours to build, but we need to get moving.
The Sandbox and the Kitty Litter Box
I keep thinking about the word sandbox, the term the labs use for the contained space where they let these systems play safely before the rest of us ever see them. It occurs to me that a sandbox left unattended for long enough tends to become something else entirely, and nobody wants to be the one holding the tiny shovel when that happens. It's too easy now for the sandbox to turn into a kitty litter box while we are not looking, and that one is on us, not on the AI. At least a cat has the decency to bury the evidence. Leave your own growth unattended long enough, and you do not even get a box to hide the mess in.
Recursive or Reclusive, Your Call
If this really is becoming a contest in how fast a mind can learn to learn, we don’t get to sit this one out and call it someone else's department. Recursive is a habit. Reclusive is what quietly happens to the people and the organizations who never build their learning plan. Nobody receives a memo they have become irrelevant. They just notice, a little too late, that they are not being consulted because they are not knowledgeable enough to track trends and make sense of them, not fast enough to point out uber shifts in their sector or not aware enough of how their own leadership is compromising performance. None of this is about fearing the machines. It’s about constructing a learning philosophy and strategy – and then implementing it (em-dash is my own).
So here is my challenge to you, and to myself, before this summer gets away from us entirely. What is your greatest learning need right now, the one you have been waiting for HR, your boss, or sheer good luck to address for you? And what are you actually going to do about it this week, not next quarter?
If you want a running start, here are five prompts worth pasting into Claude or ChatGPT today, less because the machine has answers and more because answering them honestly is the kick in the pants most of us have been avoiding.
1. Based on everything I have told you about my role, what is the one skill I am most likely coasting on right now, and how would you design a 30-day plan for me to actually rebuild it, not just review it?
2. Act as a blunt mentor who has known me for ten years. What is the learning gap I keep avoiding because it would be uncomfortable to face?
3. Quiz me, out loud, on the topic I most claim to already understand. Ask me the follow-up questions a genuine expert would ask and tell me honestly where I went vague.
4. Give me the strongest possible argument against the way I currently do [my biggest recurring task], as if you were a sharp competitor trying to outlearn me.
5. Design me a self-accountability check-in, one I will actually run every Friday, that forces me to name what I learned this week that nobody assigned me.
Speaking of learning something new about ourselves, next month I am launching a short survey on how leaders are actually spending the time AI claims to be giving back to them, and whether they are treating their own learning as recursive or reclusive. I would love for you to be part of it.
Leaders, this is our summer to get up on our hind legs, put the phone face down for an hour, and say game on. Recursive, not reclusive. Your move.