A newspaper of my own
What happened when I started using AI to curate my reading instead of scrolling for it.
Does anyone else feel less and less control over what they see every day? Algorithms govern every single major social platform, and even ones that claim to be more reader centric and less about advertising have started incorporating some algorithmic discovery and feeds for users to scroll. Finding nourishing content to consume is a real challenge in the current information landscape. Whether due to AI generated content drowning out authentic writing, or due to algorithms feeding us stuff that we’re not actually looking for, I’m finding it a lot harder to discover new, impactful, and thoughtful writing when I sit down with the intention to read a few articles during lunch.
The consequences of an uncontrolled digital diet aren’t very widely discussed. It’s easy to find information, tips, and advice on our eating habits and diets. The internet is full of discourse about the best diets, but we spend much less time discussing what we should be feeding our minds. I suspect that managing what we feed our minds was easier to do when the only option was physical media, but as technology has continued to advance, it has become much harder to be intentional about what we consume. Information diet is something that is so important since what we read, watch and consume affects what we think, which has a direct impact on the quality of our lives.
Low quality AI generated content, AI summaries and daily digests are further issues I’ve been having with my digital diet. It feels like most of the articles I’ve been coming across on X lately are at least AI assisted, and I’ve been finding it really difficult to intentionally consume media around topics I’m genuinely interested in, especially if it’s unrelated or far from my existing interests on the platform. I want to be reading human writing, from the source directly, not a summary of a third hand retelling of a story or a reaction piece. Reading original writing is something I value and have been attempting to do more of.
Dan Koe often discusses the importance of taking control of our digital environment as one way in which we can determine the outcomes of our lives. I’ve been thinking about how we practically do that in this day and age. With social media platforms being the primary digital places that most people consume information on, algorithms that we have no insight into and limted control over govern what we see and thus, what we think. Some people are able to train their algorithms to show them what they want to see, and manage to avoid content that doesn’t serve them, but the awareness and focus required for this is beyond most people, certainly for me. These platforms are designed to draw your attention, and it’s so easy to forget what you wanted to do on the app in the first place. They have no incentive to allow users to make things quieter, or to consciously determine what they encounter. Algorithms that have been obscured from the user and made intentionally hard to deliberately alter make taking control of your digital environment extremely difficult to do.
There are existing solutions to the problems I’ve raised. Using a platform like Substack or X, you could follow accounts you know about and that are aligned with your goals. This is often a good starting point for consuming more intentionally. You could then scroll the chronological feed, but you always get that one account that posts 100x more than all the others you follow, drowning everyone else out. Exploring new interests is driven by discovery and curiosity once you have a starting point. Algorithms are excellent for providing that starting point. They allow us to easily come across new ideas that we wouldn’t normally think to search for and wouldn’t otherwise encounter. You can learn so much from following a curiosity driven rabbit hole once you’ve got a good starting point. However the algorithms should be optimising for the wellbeing and intentions of the users, not the platforms selling attention for ad revenue.
Another solution I’ve tried in the past is to separate scrolling and reading by saving content to a Read-It-Later service. This is an excellent idea, since you do the algorithmic discovery and accept that getting distracted is okay, and then save the things that you’d like to read more intentionally somewhere else, and get back to those in a dedicated time slot. I used Readwise Reader for a while and had the habit of doing this. It was a wonderful way to consume more intentionally, but I inevitably built up a massive backlog of reading I’d never get through. Opening the app became a reminder of that mountain of shame and regret so I’d opt for the easier option of finding a new article to read on Substack or X.
A potential remedy I’m experimenting with is based on the form of a daily newspaper or magazine. I have AI scour the internet for articles and the occasional video. It is instructed to compose a limited edition that is tailored to my specific interests and desires to learn, with intentional room for expanding what I encounter. Limited in scope, no pressure to read what gets presented, targeted to my stated preferences and goals, with a few pieces to introduce me to new things I wouldn’t have otherwise found. I use my existing ChatGPT subscription for a scheduled job that curates the edition daily, which contains roughly 2 hours’ worth of reading material. I’ve been extremely happy with the results after using it for two weeks and have found some gems this way that I wouldn’t have otherwise found. (This article about the joy of cutting wood, the city design of Edo in the Samurai era, and this article about the beauty of West Coast surf spots being jeopardised by coastal mining, and many more). I believe that using AI to serve as a filter for the vastly expanding digital ocean of content is an under-explored avenue, and can serve the individual much better in discovery than traditional engagement based algorithms on social platforms, as long as the AI remains aligned with your best interests, (more on that another time…). I don’t think we should be offloading the reading or understanding of content to AI, but discovery and scouring through hundreds of articles, videos and other forms of content that exist online to find the treasure among the trash (as measured by your personal interests and desires) is something that is extremely well suited to large language models.
I’ve been working on turning this experiment into a product I can share with others, and it’s nearly ready for release. Subscribe to hear more as I post updates about the build and link to it when it’s ready for the first users. I expect to have an announcement about that up early next week. Stay tuned.
Special thanks to Sean Muller for pushing me to publish this despite not feeling ready.
AI statement: I used AI to assist with the title, subtitle and for reviewing the piece. All writing is my own.