The ambition technology: why maximum AI value starts with a bigger ask, not a smaller workload

Overview

Two findings landed in the same week of July 2026. Uber’s chief technology officer described how small two-week “agentic pods” — an AI-proficient engineer paired with someone who knows the work — turned a 15-hour capital-allocation job across 150 cities into a 30-minute one. And Section’s AI Proficiency Report, built on assessments of more than 5,000 US knowledge workers, found that while 69% said their organisation had taken action on AI agents, only 16% had actually used one at work. Same technology, same month, and outcomes that don’t belong in the same economy. The variable isn’t the model.

Behind that gap sits a live argument about people. David Brooks, writing in The Atlantic in late June, argued that thriving with AI will track a psychological trait — your appetite for hard thinking. Nathaniel Whittemore, on the AI Daily Brief, pushed back with a different claim: the appetite isn’t fixed, and the tool itself can raise it — if you stop treating AI as a way to do less.

This weekly is about the frame that decides what AI is worth to you: the difference between asking it to shrink your workload and asking it to enlarge your reach. By the end you’ll have both sides of the argument, the evidence that the “work less” promise was hollow anyway, and a way to test the bigger ask on work you’d already written off.

The content

Start with the pessimistic case, because it’s serious. Brooks’s essay — subtitled “what will differentiate people is not how smart they are but their relationship to mental effort” — leans on a real psychological construct: some people have a high need for cognition and enjoy thinking hard, while “cognitive misers” take any opportunity not to. His most quotable line is the hinge: “When intelligence is plentiful, volition is valuable.” When competent output is cheap, what’s scarce is the will to keep engaging your own mind. And there is early evidence for the fear underneath it: an MIT Media Lab study — a preprint, 54 participants, essay-writing under EEG — found that people writing with an LLM showed the weakest brain connectivity of the groups tested. Small study, narrow task, but it points where Brooks points: used as a substitute for thinking, AI lets you think less, and most people will take the deal.

Read that way, the future belongs to a temperament — the people Whittemore glosses as the “mental marathoners,” who keep doing hard cognitive work because they like it. His response, in the July 12 episode, is the reframe this weekly exists for. The erosion evidence describes one way of using AI: pointing it at your existing tasks and pocketing the effort saved. But that was never where the value was. AI is better understood, in the framing he’s used on the show, as an ambition technology — its highest use is “doing things that weren’t possible before,” not doing yesterday’s work more cheaply. And the appetite Brooks treats as a trait, Whittemore treats as suppressed capacity: “We’ve under-asked of people for a very long time.” Jobs carved into narrow, repeatable task buckets trained people to make small asks — of themselves, and now of their tools.

Put concretely, there are two frames you can hold while sitting in front of the same system. The labour-saving frame asks: what can I stop doing? Its ceiling is your current to-do list — the most it can ever return is the effort you already spend, and it quietly trains the substitution habit the MIT study worries about. The ambition frame asks: what is now worth attempting that wasn’t? That question has no ceiling, and it demands more of your thinking, not less — you have to specify work that has never existed. June’s weekly on the imagination gap argued that judgement about what to point AI at is the skill that divides people. The ambition frame is the step before that judgement can even operate: whether you let yourself consider targets beyond the work you already do.

Here’s the twist the research adds: the “work less” promise doesn’t survive contact with reality anyway. An eight-month ethnography inside a roughly 200-person US tech company, by UC Berkeley Haas researchers Xingqi Maggie Ye and Aruna Ranganathan, found AI consistently intensified work rather than lightening it. Employees “worked at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day, often without being asked to do so.” People started doing work they’d previously outsourced — coding, engineering — because AI put it in reach, and “the scope of what counted as ‘my job’ widened.” The expansion is coming either way. The only question is whether it arrives as chosen ambition or as sprawl — work seeping into lunch breaks and evenings because nobody decided what the bigger ask should be.

Uber’s pods are what the chosen version looks like. Rather than handing everyone a licence and hoping, the company paired roughly 30 of its most AI-proficient engineers with domain experts in finance, legal, and HR for two-week sprints — 16 pods across 16 business functions in two months. The engineers’ job was not to automate the task list — in the AI Daily Brief’s summary of the approach, “the workflow becomes the unit of automation, not the individual task” — and that required sitting with the work itself. As CTO Praveen Neppalli Naga put it: “You can’t automate them effectively by looking at process diagrams or documentation. You have to understand how the work actually gets done.” The results — capital allocation from 15 hours to 30 minutes, financial pacing reports from two days to ten minutes — came from raising the ask to the level of the whole workflow, deliberately, with the person who knew the work in the room. Ambition, it turns out, is not a personality trait. It’s an organised act.

Two honest cautions before you carry this into Monday. First, a bigger ask without judgement just produces confident, worthless novelty at greater speed — ambition tells you to raise the target, not that every raised target is worth hitting. Second, the ambition frame is not a licence for workload creep: the Haas study is a warning that “more” will colonise your evenings by default. Choosing ambition means swapping the ask up — retiring the low-value work the tool now covers and spending the recovered capacity on the previously impossible — not stacking the new on top of the old.

If you want to test the frame this week, run an ambition audit on your own shelf. Somewhere in your role is work you decided long ago not to do: the analysis that takes a day you never have, the client-ready summary that needs skills you’d have to borrow, the review of a hundred documents you sampled at ten. Pick one, open the AI tool you already use at work, and hand it the whole job rather than a task — what the work is, why it was shelved, what “done” would look like, and an instruction to ask you how the work really gets done before it proposes a path. Then judge the result twice: first against the honest baseline, which is that you were going to have nothing, and then on its merits, because “worth attempting” is not the same as “done well.” If the output lands most of the way to something you’d wanted for a year, you have just located value the labour-saving frame could never have found, because it only ever looked at work you were already doing.

Additional reading

• How to Help People Thrive with AI — The AI Daily Brief (12 Jul 2026) — the episode this weekly draws on: Brooks’s argument, the ambition-technology counter, and the Uber, Section, and Berkeley evidence read together.
• The People Who Will Thrive in the AI Age — David Brooks, The Atlantic (28 Jun 2026) — the need-for-cognition argument in full; “when intelligence is plentiful, volition is valuable” (paywalled).
• The AI Proficiency Report — Section (dated June 2026, released 7 Jul 2026) — 5,000+ US knowledge workers assessed: 69% of organisations acting on agents, 16% of workers using them, and fewer than 10% able to define one.
• Agentic AI adoption is on fire at Uber — Praveen Neppalli Naga on X (7 Jul 2026) — the primary post: 99% of Uber engineers using AI tools, 70%+ of pull requests attributed to agents, and the pod model.
• Uber’s CTO embedded its top AI engineers in HR, finance, and legal — Business Insider via Yahoo Tech (9 Jul 2026) — the pod structure and results, with Neppalli Naga on understanding how work actually gets done.
• AI promised to free up workers’ time. UC Berkeley Haas researchers found the opposite. (18 Feb 2026) — Ye & Ranganathan’s eight-month ethnography: intensification, widening job scope, and work seeping into pauses. Published in Harvard Business Review as “AI Doesn’t Reduce Work — It Intensifies It” (9 Feb 2026).
• Your Brain on ChatGPT — Kosmyna et al., MIT Media Lab (preprint, 10 Jun 2025) — the EEG essay-writing study behind the cognitive-erosion worry; 54 participants, LLM users showing the weakest connectivity.

Editor’s note

This site is the example I can vouch for personally. A daily learning publication (researched, verified, and shipped to a schedule) was never a project I had declined for lack of time; it had never made it onto a list at all, because before agents the cost put it beyond consideration. I notice the same pattern across the rest of my week. Where I use AI to compress work I was already doing, such as a first-pass read of a contract markup, the return is real but modest. The work that was never going to happen at all is where the return is out of proportion. So when you take stock of what AI is worth to you, count what you now attempt, not just what you finish faster.