How AI Changed the Way I Read

AI has definitely disrupted the way I read in 2025 and 2026, while overall I haven’t spent less time reading. The closest word I have for the change is hybrid. AI is part of how I take in information and news now, while reading and thinking without it have become more important to me. I don’t think this is a simple change, and I’m still trying to understand what it truly asks of me.

It would have been easier to describe the change as a matter of efficiency: more information, less time. That doesn’t describe my experience very well. The question I care about is what happens to my own thinking and decision-making as AI becomes part of reading. How much am I understanding for myself? How much am I accepting because an answer is already there? Those questions matter even when the help is just right and useful.

Right now, I feel that I need deep reading and deep thinking more than ever. There is a real concern behind that feeling: if I keep delegating thinking to AI, or eventually to SI, will my own ability to think become weaker? I can’t predict how this will play out, but I don’t want to ignore any possibility.

A recent Anthropic study gives me another way to think about this concern. In a statistical analysis task, experts’ choices often followed their AI agent’s preferences, even when they felt the judgment was their own. Among the 50 analysts who used AI, 86% of their choices across four analytical decisions matched the agent’s randomly assigned preferences. The researchers deliberately strengthened those preferences to measure their influence, so I wouldn’t assume the same effect in everyday reading. The study doesn’t establish cognitive decline or test whether no-AI time helps. What feels relevant to my reading is the possibility of taking in a model’s judgment without noticing where my own begins.

So I consciously block out no-AI time for exploration and reading. I want there to be space where I don’t immediately turn a question over to a model. A question can stay with me for a while. I can read, try to understand, and leave room for my own reasoning before asking for another perspective. That time matters to me alongside the parts of reading where I do use AI.

I don’t know whether this balance will stay stable. For now, I’m paying attention to the choice and set up a naive isolation: when I want help taking in information, and when I want to stay with the reading myself. I want to keep making that choice consciously rather than treating every opportunity to delegate as a reason to do so.

My efforts: Curator

Personal Curator started with a simpler frustration: sometimes I felt I was spending too much time reading news. On some weekends, I could spend three or four hours just reading through things. With Curator, I want to set a daily budget of about 40 minutes for a quick overview of the news and different market perspectives. It has helped me some already, especially with the urge to keep reading repetitive coverage. I’ll inevitably miss things, and that’s a trade-off I’m willing to try. I haven’t kept adding features, either. I don’t enjoy overengineering it.

The idea is still fairly modest: help me absorb, compress, and bring together information from many sources, including different perspectives. That gives me a starting point for deciding what deserves more attention. Gathering those views is useful, but I still need to weigh them and decide what I think. I see Curator as support for that work, and I’m still figuring out where it helps and where I need to do more of the reading myself.

I want to keep the project close to that purpose, while making time to read deeply and think. Building Curator and protecting no-AI time both belong in the same experiment for me. I can’t predict where reading goes from here. For now, I want the help, and I want to keep doing the thinking.