Random Work Is the New Deep Work: Notes on the IQ → EQ → AQ Shift
My work journal writes itself: every session gets captured by a hook, and a nightly job distills the day into one page. Eighty-four entries since April 4. This morning I scrolled back through all of them looking for a through-line, and the honest answer is there isn't one. An EDI acknowledgment loop on a Monday. A drift detector for model routing on a Wednesday. License hygiene on a Friday. A kernel-panic postmortem the week after. One hundred fourteen article seeds sit in my backlog right now, and they read like ten different people wrote them.
The voice in my head about this is the one every career book installed: pick a lane. Depth wins. A senior engineer is someone who spent a decade getting unreasonably good at one thing. By that standard, the last two months of my life look like a focus failure.
I want to argue the opposite. I also want to be careful doing it, because the argument flatters me, and arguments that flatter you are the ones to check twice. So I treated it like any suspicious result: I went looking for the data that would kill it.
Quotients have a habit of taking over
For most of a century, the prized number was IQ. It sorted school tracks, army placements, and eventually, through a long chain of proxies, who got hired to think for a living. The assumption underneath it: intelligence is the scarce input, so measure the intelligence.
Then in 1990 two psychologists, Peter Salovey and John Mayer, defined "emotional intelligence" in an academic journal. In 1995 a science journalist named Daniel Goleman turned it into a bestseller, and that October TIME put EQ on its cover. Within a decade every leadership offsite had a module on it. EQ also got oversold: the popular claim that emotional intelligence accounts for up to 90 percent of leadership success never had adequate data behind it, and by 2005 the psychologist Edwin Locke was publishing papers calling the whole construct invalid. Hold onto that pattern. We'll need it again.
Adaptability's turn started quietly, at the level of companies rather than people. In 2011, two BCG strategists argued in Harvard Business Review that sustainable competitive advantage was dying and the replacement was speed: the ability to read signals, experiment, and mobilize faster than the environment changes. By 2019 a venture investor named Natalie Fratto was on the TED stage proposing AQ, the adaptability quotient, as the thing she screens founders for. An assessment industry followed, the way it always does.
Here is where I'm supposed to quote Darwin about how it is not the strongest of the species that survives but the most adaptable. He never wrote it. A management professor named Leon Megginson paraphrased him that way in a 1963 speech, and the paraphrase got promoted to scripture because it was too useful to fact-check. I find that fitting rather than damning: the adaptability era runs on a quote that adapted.
You don't need Darwin. You need job ads.
Here's what I found when I tried to kill the thesis.
LinkedIn's learning report had already named adaptability the "skill of the moment" in 2024, the fastest-growing skill demand in its data. The World Economic Forum's employer survey expects 39 percent of core skills to be transformed or obsolete by 2030; the previous edition said 44 percent by 2027, so the panic cooled a notch while the direction held. PwC mined close to a billion job ads and found the skills employers ask for changing 66 percent faster in AI-exposed occupations than in the rest of the economy, up from 25 percent faster one edition earlier. The churn isn't spread evenly. It concentrates exactly where builders live.
Then the stat I keep rereading. When Microsoft and LinkedIn surveyed 31,000 people in 2024, 71 percent of leaders said they'd rather hire a less experienced candidate with AI aptitude than a more experienced candidate without it. Read that again slowly. Experience is the compound interest of the IQ era, the asset you were told to spend thirty years accumulating. A majority of hiring managers just said they'll trade it for evidence you can absorb a new tool this quarter.
The quietest data point comes from inside psychometrics itself. For decades the textbook said general mental ability tests were the single best predictor of job performance. In 2022, Sackett and colleagues re-ran the math and showed the classic estimates had been systematically over-corrected for years. In the revised table, cognitive ability tests fall behind structured interviews and biodata. Demonstrated behavior now outranks measured aptitude in the discipline that engineered aptitude measurement. Nobody held a parade. The most-cited number in hiring science got quietly marked down, in the same decade the market started pricing adaptation.
The people building the tools say it in plainer words. Jensen Huang stood on a stage in Taipei in 2023 and declared, "Everyone is a programmer now. You just have to say something to the computer." Sam Altman keeps answering the what-should-students-learn question with versions of one answer: learning how to learn, resilience, the raw ability to adapt when everything around you changes. Dan Shipper calls what comes after the knowledge economy the allocation economy: you stop being valued for what you know and start being valued for how well you direct intelligence that isn't yours.
So the data didn't kill the thesis. It sharpened it.
What it feels like from inside
What the reports can't tell you is what the shift does to the person living it.
My family farms. They do work that pays off only when the season turns, and I inherited that patience along with the assumption that mastery has a season too: plant the years, harvest the expertise. The hardest thing about building with AI is watching my field lose its seasons. IBM's researchers put the half-life of a technical skill around two and a half years now. From inside, it feels shorter. Frameworks I knew deeply stopped mattering. Tools I dismissed became load-bearing within a quarter. The capital I'd spent a career compounding was melting while I held it.
I can date the low point. On May 13, I kernel-panicked my own machine: one local model too many pulled into memory while another heavy job was already loaded. The computer that runs my whole operation went dark because I was trying to absorb new tools faster than I was respecting their limits. That's the texture of this era that never makes the keynote: the am-I-keeping-up loop, the vertigo weeks, the retraining that happens at hours the journal timestamps don't flatter.
Here's the part that made me stop reading the crash as a verdict. By the end of that same day, the panic had become two new entries in my article backlog: one on queueing disciplines for local model fleets, one on postmortems for solo builders. Both seeds carry the source date May 13. The crash and its curriculum, logged on the same page.
That's when the journal's through-line finally showed itself. I'd been scanning the topic column, and the topics never repeat. The pattern lives in the other column, the one that never changes: frame the problem sharply, find the prior art, set the quality gates, put the machines to work, audit what comes back, write down the lesson. Every one of those eighty-four days runs that loop. An EDI acknowledgment protocol and a drift detector have nothing in common as domains. As loops, they're the same day.
Once I saw the loop, the economics flipped. A new domain used to cost months of ramp before output; now an unfamiliar one goes from hostile to workable in about a day, because execution is cheap and the loop is practiced. Seventeen days after the kernel panic, an essay on drift detection for model routing went out the door. Watching working code materialize in a domain I didn't know the week before is the closest thing to a cheat code I've ever felt while building. The dopamine is real. So is the discipline bill: the faster the code appears, the more the verification matters, because speed without gates is just confident garbage.
That loop is what people are trying to name when they say AQ.
The honest caveats
I'm reluctant to turn this into a score, because the science isn't there. AQ has no validated instrument the way IQ does; the prominent assessments are commercial products; and the measurable parts of adaptability keep dissolving into older constructs when researchers look closely. A 2017 meta-analysis found Big Five personality traits explain a large share of what career-adaptability scales capture. Awkwardly, "learns fast in unfamiliar situations" was always half of what intelligence tests measured anyway. The quotient framing is marketing. EQ taught us how that movie goes.
But the repricing is not marketing. The job ads, the hiring preferences, the skill-churn rates: those are measured behaviors of people spending money. You can reject AQ as a construct and still accept the conclusion that the market premium moved from what you've mastered to how fast you re-orient.
Recovery time, not mastery time
So here's the principle I run on now. I stopped optimizing time-to-mastery in a single domain and started optimizing recovery time across domains: the interval between landing somewhere unfamiliar and shipping something I can defend there. It's days now. I'm working on hours. You can't put that number on a résumé yet. You can only demonstrate it, which may be exactly why the selection methods that survived the 2022 revision are the ones that watch you behave instead of asking what you know.
The deep-work instinct isn't wrong; it just attached to the wrong object. The thing to go deep on is no longer a domain. It's the loop that eats domains.
Eighty-four journal entries, one hundred fourteen seeds, one kernel panic, and no through-line in the topic column. I spent two months reading that as the bug.
It's the résumé.
Top comments (0)