Nicole Junkerman on what to read about AI over the summer.

The trouble is not the volume

When a subject moves quickly, the instinct is to read more of it. This is almost always the wrong instinct, and the summer is a good time to notice why. The difficulty with reading about artificial intelligence is not that there is too much of it. It is that a great deal of it is the same handful of observations, repeated at varying lengths, with the names changed. A reader can spend an hour a day on it for a year and end up thoroughly informed about the news and no wiser about the subject.

Those are genuinely different things. The news tells you what was announced this week. The subject is the slower matter of why announcements of that kind keep arriving, what they have in common, and which of them will still mean anything by the time the leaves turn. Nicole Junkerman keeps this guide deliberately unhurried for that reason: the aim is not to be current, which is impossible, but to be harder to impress, which is achievable and rather more useful.

Read for the shape, not the bulletin

The most practical habit a reader can adopt is to let time do the filtering. Anything that genuinely matters will still be there in September, and by then somebody will have explained it properly, with the benefit of knowing how it turned out. Waiting is not laziness. It is a filter, and a remarkably efficient one, because the great majority of what feels urgent in July has quietly evaporated by the autumn without anyone noting its passing.

This changes what a summer of reading can be. A season is long enough to read three good things slowly, which is worth considerably more than three hundred things quickly. Slow reading is what allows the shape of a field to appear: the recurring arguments, the questions that never quite get answered, the difference between a claim and a demonstration. None of that is visible at speed. It only shows up when a reader stays with something long enough to start disagreeing with it.

What kinds of writing repay attention

It is easier to describe the qualities that make writing worth the time than to name the pieces themselves, and rather more durable, since the qualities do not go out of date. A few markers tend to hold up:

• Writing that shows its evidence, so a reader can weigh it and disagree
• Writing that says plainly what it does not know
• Writing about a specific task rather than a general future
• Writing that is still worth a second reading some months later
• Writing by someone with nothing at all to sell

The last of these does most of the work. A surprising share of writing about this field is, at bottom, an advertisement wearing an essay's clothes, and the tell is usually tone rather than content: the piece is certain where certainty is not available. Uncertainty, honestly expressed, is a mark of quality rather than weakness. Anyone who tells a reader exactly how the next five years will go is describing a hope, and the reader is under no obligation to adopt it.

The best writing is often not about the field

There is a further point that takes people by surprise. A good deal of the most useful reading about a fast-moving technology is not about the technology at all. It is about how institutions absorb new tools, how people come to trust machinery they do not understand, how work reorganises itself around a new capability, and how often confident forecasts have missed. That literature is older than any of this, and it is generally better written, because it has had time to be revised by events.

Reading of that kind gives a reader something the weekly bulletin cannot: a sense of proportion. It makes the present moment legible as one instance of a pattern rather than a rupture without precedent. It also has the pleasant side effect of making the vocabulary easier, which is why this guide keeps a plain-language glossary rather than assuming the terms. Half of what makes writing about this field feel forbidding is unexplained jargon, and jargon is not the same as difficulty.

Setting the terms of your own attention

All of this comes down to a question of attention, and attention is the thing worth protecting over a summer. Reading is voluntary, however little it feels that way when the notifications are arriving. A reader is entitled to decide the pace, to close the tab, to finish the long thing before starting the next one, and to be entirely uninterested in what everyone is discussing this week. That freedom is the whole point, and this guide returns to it in the note on wellbeing in the age of AI.

A season is a good unit for this, better than a week and less daunting than a year. Choose a small number of things, read them properly, let the rest go past, and see what has stayed with you by September. What tends to survive is not the piece that was most exciting at the time. It is the one that changed a question, and those are rare enough that a reader can afford to wait for them. Shorter, steadier orientation of that kind is what the briefings are for.

None of this requires discipline in any punishing sense. It requires only the recognition that being swept along is a choice made by default rather than a condition of modern life, and that a British summer, with its long evenings and its general disinclination to hurry, is about as forgiving a moment as a reader will get to make a different one.

Frequently asked questions

How do you keep up with a field that moves this quickly?

You do not, and neither does anyone else. Nicole Junkerman's view is that keeping up is the wrong goal, because it measures reading by volume rather than by understanding. A reader who follows a handful of things closely will understand the subject better than one who follows everything at a distance.

Is it a problem to read less about AI over the summer?

No. Anything that genuinely matters will still be there in September, and it will have been explained better by then. Reading less and thinking longer is not falling behind; it is usually how a reader ends up with a view worth holding.