I have been saving reading highlights for about four years. A sentence stops me, I mark it, and it goes into an archive I almost never open. There are several hundred of them now.
Last year I asked the software what they meant. Specifically, I asked what my highlights said about living a worthwhile life.
It gave me nine themes. Challenges are energising. Do more of the activities you actually choose. Failure is a stage rather than a verdict. Fear is mostly the belief that you cannot cope. Start from the body. Collect experiences. Authenticity needs responsibility attached to it. Boundaries protect your energy. Stop asking what life means and get involved in something.
Each theme came with a practice. Notice a challenge you would usually avoid. Schedule twenty minutes of something that nourishes you. Break the avoided task into its smallest step.
I published it. It has been on this site since.
Nothing in it is wrong. That is the part worth explaining, because my first reaction was that the answer was rather good, and my second, some weeks and months later, was that nothing had happened as a result of reading it.
I have been trying to work out why.
One possibility is that a highlight is not a belief. It is a record of a moment when something got through. The reason I saved a particular sentence is almost never contained in the sentence. I marked a line about fear because of a week I was having. I marked a line about novelty because I was somewhere unfamiliar and it fitted. The software had the sentences and none of the occasions.
Another is that the AI synthesis flattened four years into a single voice. The person who highlighted the line about fear was not in the same state as the person who highlighted the line about collecting experiences, and they might not have agreed with one another. Averaging them produces a philosophy that nobody actually holds, which may be why it was so easy to nod along to.
There is also the format. Every theme arrived with something to do. I had not asked for that. It came because this is what the genre requires: insight, then instruction. The highlights did not imply any of those practices. The shape did.
The explanation I keep returning to is less comfortable than any of these.
Saving the highlight was already the move. Something in a book lands, and there is a small decision about what to do with it. I can stay with it, which is uncertain, takes time and may produce nothing. Or I can mark it, file it, and feel that it has been dealt with. The archive is not neutral storage. It is a way of converting being affected into having an item.
Asking a machine to summarise four years of that is the same operation, performed faster and at scale. It produced what the method was always going to produce: nine statements that are sensible, complete and inert.
I recognise the shape from elsewhere. People sometimes arrive in therapy with a full account of themselves already assembled. The childhood, the pattern, the terminology. Nothing in it is inaccurate. It has also not changed anything, and it can take a while to notice that the account is doing a job, which is to keep the material at a workable distance.
I do not think the answer is to abandon the tools. I still save highlights. I have stopped treating the collection as raw material for conclusions.
What it is good for is narrower than I thought. It is a record of several hundred occasions when a sentence stopped me. Read that way, the interesting question is why that one, on that day.
Four years of highlights turned out to contain nine good ideas.
The same four years are a record of several hundred moments when something got through to me.
For a long time I was reading the first one.