A weekend long-read on the human side of AI: David Brooks says only the mental marathoners survive the AI age — NLW pushes back, arguing the real opportunity is using AI to do things you couldn't do before, with Uber's Agentic Pods as the proof of concept.
The people who thrive with AI won't be the ones who work less — they'll be the ones who use AI to do things they couldn't do before.
Brooks argues that in the AI age, what differentiates people is their relationship to mental effort, and that only the "mental marathoners" who resist AI will thrive. NLW agrees the efficiency framing is a trap but disagrees with the fatalism: the biggest opportunity isn't shaming people for offloading rote work, it's using AI as an ambition technology to stretch capabilities and reinvent work. Uber's Agentic Pods show the model — pair AI-proficient engineers with business experts, and the real gains come not from the productivity itself but from what people do with the reinvested time.
Section's latest AI proficiency report found that while 69% of workers said their organization had taken some action on AI agents, only 16% actually use an agentic tool at work and fewer than 10% can define an AI agent in their own words. Only 30% of employees at agent-using organizations have received any agentic training.
The week was dominated by model releases, but NLW argues model improvements are 'all for naught' if people aren't supported in learning how to extract value from them. Capability without adoption support is wasted.
Citing Brooks, an ActiveTrack analysis of 10,000+ workers found that adopting AI more than doubled time on email, messaging and chat apps and raised business-software use 94%. Focused, uninterrupted work fell 9% — a state now nicknamed 'AI brain fry.'
UC Berkeley Haas researchers found workers began taking on tasks they'd previously outsourced — coding, engineering — because AI made them easier, squeezing in work bursts on evenings, weekends and in waiting rooms while multitasking across many bots. The saved time gets spent on new tasks.
— David Holz, Midjourney founder, on X. Midjourney founder David Holz tweeted that the productivity-plus-exhaustion feeling made him think 'something is wrong, and also that there might be a big opportunity,' asking for day-to-day strategies to make it feel better.
NLW revisits his 'infinite backlog' idea: because agents don't sleep or take weekends, it feels like there should never be downtime. In reality the limit hasn't disappeared — it's shifted from how much we can do to how much planning and oversight we can support.
— David Brooks, 'The People Who Will Thrive in the AI Age' (The Atlantic). Brooks argues the people who make a difference won't be those who use AI to work less, but those who actively wrestle with it to develop their own capabilities. What differentiates people, he says, is not how smart they are but their relationship to mental effort.
Brooks divides people by their psychological 'need for cognition': those who enjoy thinking hard, cognitive misers who avoid it, and a medium group in between. It correlates with intelligence but isn't the same — plenty of smart people dislike hard work.
Brooks's first archetype has a low need for cognition and uses AI to do less. AI still helps them get more done, but the risk is that by making tasks easy it diminishes their capabilities.
Brooks cites MIT Media Lab research finding people's brain connectivity declines as much as 55% when using ChatGPT versus performing similar tasks without it, plus a Possibility Sciences study showing gamma-wave activity — a marker of cognitive effort — dropped roughly 40% with AI. He fears predictable erosion of critical thinking.
People with a medium need for cognition understand AI might hollow them out and resolve not to over-rely on it — but under everyday stress their resolve fails. The core danger, Brooks says, is optimization: chasing maximum output instead of excellence.
In a survey for software firm GoTo, 43% of workers said they had submitted AI-generated content even though they suspected it contained errors and was generally low quality — a symptom of optimizing for output over excellence.
— Chris Sibon, head of school at Rivendell, via Brooks. A school head, Chris Sibon, showed students a film that took 200+ artists five years to make; students asked why, since 'AI could have done it in five minutes.' His point: a student who wrestles with a hard text and fails and tries again is more than informed — he is more solid.
Brooks's third archetype are the high-need-for-cognition people who run mental marathons the way some run 26.2 miles despite the car existing. In the AI age he suspects they'll work hard to resist AI, wanting original, personal work and using AI to increase agency rather than diminish it.
— David Brooks, The Atlantic. Brooks's more hopeful turn: need for cognition isn't a fixed trait — willpower is 'extremely sensitive to context.' The crucial task, he says, is to cultivate people's desire to seek out cognitive complexity, so AI does the calculating while humans define what matters.
NLW's central disagreement with Brooks: rather than shaming people for offloading rote work like functional emails, the real power is doing things that weren't possible before. People whose brains are 'lighting up' aren't resisting AI — they're using it as an ambition technology to stretch their capabilities.
NLW describes the humbling loop of a non-coder building an agent — asking AI how, screenshotting the answer to ask another AI what it means, hitting errors, shipping something that crumbles on first contact. Like physical workouts, mental elasticity comes from doing what's uncomfortable; AI hasn't changed that, it's raised the ceiling on ambition.
The WSJ CIO Journal profiled 'AI champions' — super-fans who get early tool access and training in exchange for promoting adoption and fielding colleagues' questions. One law firm is formalizing a program around ~60 champions to track and scale their impact.
Where the WSJ framing misses, NLW argues, is treating champions as internal PR agents. The real value isn't telling people how good AI is — it's showing them what they could actually be doing if they tried.
NLW's 2026 prediction: mirroring the forward-deployed-engineer trend, companies will grow internal builders who pair with business functions not to make current work 20% faster, but to fundamentally change how — and even what — those functions do.
Uber CTO Praveen Nepali described pairing ~30 AI-proficient engineers each with a domain expert for a two-week sprint: days 1-2 shadow the expert, day 3 prioritize, days 4-5 build alongside the worker, days 6-9 validate with peers, day 10 ship. Sixteen pods ran across sixteen functions in two months.
Per Nepali, 99% of Uber engineers use AI tools, over 70% of pull requests are attributed to local or cloud agents, and engineers have built 2,500+ agent skills across the software development life cycle. Pod results: capital allocation across 150 cities went from 15 hours to 30 minutes, financial pacing reports from two days to 10 minutes.
— Praveen Nepali, Uber CTO. Nepali's biggest lesson: the largest wins come from rethinking whole workflows — eliminating handoffs, removing approvals, replacing legacy tooling — not automating single tasks. And the best opportunities are invisible from outside; you find them by sitting next to the people doing the work.
NLW argues the two-week productivity wins are just low-hanging fruit. Over months, the locus of change should shift from engineers to business people who, influenced by agentic working, reinvent their work — spending the freed-up hours on new, orthogonal work that was 'always dreamed but never possible before.'
NLW's rebuttal to Brooks's fatalism: he suspects Brooks doesn't really believe the marathoners aren't the only survivors. In both education and work we've given people discrete task buckets for inscrutable reasons — if we support AI adoption well, we'll find far more untapped human potential than most people realize is there.