I have always enjoyed sharing my readings. Back when I was more active on Twitter, I used to share almost every post I came across. Since transitioning to Bluesky as a replacement, I’ve continued sharing links there, though I haven’t been particularly active on social media lately.
While I’ve reduced my activity on social media, Readwise’s Reader has been instrumental in helping me stay organized and focused. It’s shifted my attention toward reading and writing on my own platform rather than sharing on social networks.
For over a year now, I’ve made it a habit to share a curated list of links with my team every Friday, readings I believe they might find valuable, usually accompanied by a very short summary generated with GPT via Readwise’s Ghostreader. Fortunately, some of my colleagues have adopted this practice too, turning Fridays into a sort of collective digest where we exchange links worth highlighting. These digests have become a great way to stay informed as a group.
Initially, this process was manageable since I was only sharing a handful of links each week. All I had to do was copy and paste the links from the Recently Highlighted tab on Reader, along with the short AI-powered summaries and a few personal comments. But as I’ve been diving deeper into various topics lately, curating these lists has started to feel a bit overwhelming. While searching for ways to streamline this process, I came across Simon Willison’s approach to running a link blog. Inspired by his setup, I decided to take things a step further and create my ownlink-blog here on this site.
As noted in my AI usage statement, I use AI tools to help improve my English writing. Now, I’m incorporating AI into this new project to automatically summarize my recent readings. I hope this will make it easier for you to discover insightful reads while allowing me to share them more efficiently.
To make this happen, I developed Link Blogger a small Python program that uses a Large Language Model to classify articles by topic and summarize the highlights I’ve underlined, conveniently wrapping everything into a Markdown draft file. The input articles are documents synced from Readwise directly to my laptop.
For now, the process remains semi-automated. Every Friday, I review the draft generated by the program, add personal notes where needed, and then share it with all of you. You’ll find these posts by following thelink-blog tag.