What's The Point to Save Articles You’ll Never Read?
View original- best
Summary (TL;DR)
The author explains that saving links feels productive but leads to guilt and unread piles. After reading 40% of what they save, they shifted from saving to filtering. Their system tags incoming links as 'tbf', uses AI to summarize and compare against written preferences, and only shows matches for final selection. The key is moving the 'is this worth reading' decision to an automated filter, not reading faster. The author is building this into a tool called Readplace, but the core idea works with any AI and a preferences list.
Hoarding links feels like progress. You find something good, you save it, and for a second you feel on top of things. Then you never read it.
Most of us save far more than we read, and almost none of us have a system for choosing what to open. The pile grows, the guilt grows with it, and one day you archive the lot and start over, and they may never surface again.
I read and share a lot of writing. Around 350k Reddit karma over 11,850 submissions, and 15k on HN over 2,700, all within the decade since 2016. On average I read 40% of what I find, and I share 60% to 70% of what I read. That figure has changed over time, with more links coming in and better curation. For ten years I ran my own reading on Gmail filters, DynamoDB tables, and a stack of Reddit automations. That pipeline taught me the bottleneck was never saving. It was deciding what not to read.
Early on, once I realised a day only has 24 hours excluding sleep time, I stopped reading my pile and started filtering it.
Saving an article and reading it are different acts, one triggers instant dopamine reward and the other takes actual effort and learning.
Here’s the the system I use today, which runs on localhost:
- A newsletter and links I mail to myself land to my email and I tag them as "tbf"
- I run a UI tool that reads all the tbf emails "is:unread label:tbf" and crawl all the links on each email
- Then an AI summarises each link and checks each one against the reading preferences I’ve written down and decides whether it matches them or not.
- The one that match are listed to me for filtering, where I decide if I want to read or not.
- What reaches my reading time is already the stuff I’d have chosen, without me skimming a single headline to get there.
- Eventually some of those are posted to Reddit/HN/etc.
The trick is moving the decision, not speeding it up. You’re not reading faster, you’ve handed the “is this worth my time” call to something that knows your preferences. If you lead a team as an Engineering Manager, it’s the same instinct behind a good backlog prioritisation exercise. You don’t work every ticket, and you shouldn’t read every link.
You can build a rough version today. Write down what you actually care about and what you always skip. When links come in, have a model like Claude summarize each one and judge it against that list: “Given these interests, is this worth my time, and why in a few lines?” Only the ones that matche it can reach you. A saved prompt or a small script can run the whole pass at once.
It takes effort and trial and error to build this, to know the right prompt and the automations, which are very specific to each Engineering Manager, so I’m building it into a publicly accessible reader called readplace. The browser extensions, iPhone app, and the summaries work today, and the part that filters based on your preferences is what I’m adding next if there's enough interest. You don’t need any of it to start, because the filter is the idea, and it works today with a primary excerpt and TL;DR summary.
Saving was always the easy half. The skill worth building is teaching yourself what’s actually worth reading, and then trusting it to skip the rest.
The important bit is not what you read, but what you DON'T so you can optimise your time and, by consequence, speed up your personal development.
