The reluctant artist

One of my toddler’s favourite books is Richard Scarry’s What Do People Do All Day.

Over 60 or so pages, you follow the people of Busytown - from train drivers, to doctors, to sailors - as they go about their day.

It’s a profoundly pro-work book: one section - titled “everybody is a worker” - contains an explanation of the circular capitalist economy that would have made Adam Smith proud.

But the thing I love the most about the book is that it strips what work is back to its very core: something an individual person gets up each day, and does, to occupy their time.

It’s the thing we do all day.

Despite this, when we talk about AGI and work, we generally focus on large macro-trends. How much more productive will it make us all? What industries will be displaced entirely?

Turn to the very small: what we’ll actually be doing all day whilst living in a world of abundant intelligence, and - unless you think AI will lead to the end of work in its entirety - the answers generally aren’t here.

But they’re coming. And the first data-point came last week - via a new HBS study of the impact of GitHub Copilot on the way 187,000 developers approached their work.

Put simply, the study compared developers who used Copilot, with those who didn’t. The boring bit of the study is that it made them more productive. Well, durr.

The not-boring bit of the study is how they became more productive.

First, they farmed out the (mainly) boring project management stuff to AI. Second, they reduced the amount of meetings and conversations they had with other developers.

The result? The developers using copilot were better able to focus on doing what they valued most: more coding.

Even more not-boringly, the subjects of the study didn’t just use their extra time to do more of their core coding: they instead used it for experimentation and exploration (most notably, in this case learning new programming languages).

The humans keep the heavy intellectual lifting to themselves, and leave the AI to focus on the fringes.

This is a more optimistic world than one we often see sketched out - where we farm out the big stuff to the smartest machine in the room, limiting the role of humans to instructors and tinkerers.

Because after all, who are we to compete with superintelligence?

But this study gives us a glimpse of the alternative world: one where intelligence is not only abundant, but - for the most part - boring.

It kills tedious work, frees up time on the fringes, and leaves us to focus on the important stuff.

Take what you’re reading right now.

I do the first draft on g-docs, redraft it a few times, share it with one or two other people, and then, and only then put it through Claude for a quick check.

Alongside that, I might use Claude as an interlocutor to discuss an idea (Nate Silver described GPT in the credits for his last book “as his creative muse”), but that’s it.

It’s part of the process, but, for this kind of writing at least, it only really works if you start with human-made, and then use AI to clean up the fringes.

It’s why the approach of our 187,000 devs, feels like something that’s not only a blueprint to doing great coding, but a blueprint to doing great work, full stop.

More time spent on the core craft, more time spent exploring and experimenting, and limiting time spent collaborating to the most intentional moments.

Strip the rest of this note away, and the summary I’ve just provided reads more like the kind of advice a musician would give to an up and coming star, or a seasoned artist to a student graduating from the Royal College of Art.

Not the kind of advice a manager would give to your traditional employee in a white collar job.

What our Github developers have realised, and are really using AI to do, is to strip the theatre and unnecessary elements of knowledge work (admin! meetings!) away and instead treat their work like a true creative would.

Because they’ve figured out where their remaining value lies - creativity and craft - and used the tools available to help them do even more of that: even if they never saw themselves as creatives in the first place.

Not an artist, but working like one instead.

This way of working - the way of the reluctant artist - is not only a better way of working, but a more resilient one as well.

Even as the tech gets better, we’ll likely remain zealously protective of our most needle-moving work: prioritising the boring and mundane to automate away, and keep the best work for ourselves.

Because, for a lot of things at least, we still deep down want the human touch. And, more importantly, other humans do so as well (for a great essay on this, see James Marriott here).

It might seem naive to see this as something that will impact the entire labour market, not just founders and executives.

But here comes the final not-boring bit of the study: it found that the effects of the above were much greater for the least talented programmers, than it was for the most.

Meaning that this too might be one of those rare areas in which a rising tide does, in fact, manage to lift all boats.

So - with this first glimpse of what people will do all day in a world of AGI - it probably turns out that Richard Scarry will still have got it right.

Everybody will still be a worker, just a different kind.

aled@ashore.io