When the details get cheaper, think bigger

We have a habit of measuring our usefulness by the work we know how to do. The spreadsheet we can build. The report we can write. The complicated process we have learned to navigate.

When AI starts doing parts of that work, the question becomes personal: where does that leave us?

I think it gives us a reason to raise our sights. As more of the details become practical to delegate, we can spend more of our thinking on what we want to accomplish, why it matters, and how the pieces should fit together.

Venkatesh Rao offers a useful starting point in his October 25 essay, “Cognitive Mercantilism”, which The AI Daily Brief discussed on November 1. He challenges the assumption that we have a fixed amount of useful thinking to do. If AI makes cognitive work cheaper, he argues, people could develop new ambitions and new demand for that work. His historical analogy is an invitation to imagine that possibility, rather than evidence that a happy transition is guaranteed.

The part I want to build on is what that could mean for how we think.

“Higher levels of abstraction” sounds like something from a management seminar. In practice, it means stepping back far enough to see the purpose of the work. A formula is a detail. Deciding what a budget should make possible is a larger question. A page of software is a detail. Deciding which problem a tool should solve, and whose life it should improve, is a larger question.

Those larger questions are available to anyone who understands a problem worth solving.

Imagine a volunteer organizing rides for older neighbors. They might picture a better way to match drivers with requests, but abandon the idea because creating forms, organizing information and building a simple tool feels beyond their skills or available time.

AI assistance could make a first version worth attempting. The volunteer could describe the need, work through a draft process and try a prototype. Their attention could go toward questions that require knowledge of their neighbors: Who is being missed? What happens when a driver cancels? Can someone request a ride without a smartphone?

That is a hypothetical example, but it captures the opportunity I find most compelling. A person can bring an understanding of a need and get help with some of the machinery required to address it. We should be asking what becomes possible for people who previously could not afford that help.

There is still work involved in knowing when delegation has succeeded. A convincing answer can conceal a bad assumption. Stepping back to see the larger purpose should make us more deliberate about which details deserve close attention.

The evidence also gives us a reason to measure results. In a randomized study published in July 2025, METR found that 16 experienced developers working in familiar open-source projects took 19% longer with the AI tools tested, even though they believed they were faster. That was a particular setting with early-2025 tools. METR’s February 2026 follow-up reported that selection and measurement problems prevented a reliable estimate of the newer tools’ productivity effect; the researchers thought greater speedups were likely, but could not establish their size.

For someone trying AI, the practical question is whether the whole job improves, including checking and fixing the result. For a new project, it is also whether help makes a worthwhile attempt feasible at all.

There is a broader question about who gets that opportunity. Access costs money and time. People need room to learn, and workers facing disruption need more than advice to think bigger. We should judge institutions by whether they help people gain capability and share in its benefits. An expanding set of possibilities offers little comfort if you cannot reach it.

Still, I find real hope in the possibility that more people could act on ideas they currently set aside. A teacher could try a learning resource. A neighborhood group could test a service. Someone with years of practical experience could explore a solution without first mastering every technical step.

As AI becomes better at handling details, we should practice asking larger, more ambitious questions—and learn how to carry their answers through to something useful. What have you wanted to attempt that has always felt too complicated to begin?