I learn quite a bit. Or not less than, I save quite a bit: largely weblog articles and newsletters. I have been a read-it-later particular person because the Google Reader days, and Readwise Reader is the app I’ve appreciated greatest since then.
However someplace round article 200, I seen I had stopped studying inside it. It had develop into an awesome deposit for articles that I by no means bought round to studying. I had gotten excellent at saving and really unhealthy at ending.
And once I did end one thing good, I wished to share it, however that half was getting tough too. I had two choices:
- use the Buffer Chrome extension to queue up a naked hyperlink, which was quick however felt flat as a result of I wasn’t including any context, or
- depart the article mid-read, open a brand new tab, write one thing considerate, and lose my place within the piece.
Neither felt proper, so more often than not I simply did not publish in any respect.
Till I vibe-coded the app I wished to make use of. It is known as Reader Chomper: a small native macOS app that stitches collectively three instruments I used to be already paying for. Readwise Reader is my article backend; an LLM handles the summaries, and the Buffer API queues posts to LinkedIn and Bluesky.
The transient I gave myself was easy: a bite-sized option to learn, and a frictionless option to share.
This is the way it works.
The 4 steps
Reader Chomper does 4 issues, on this order: triage, fast scan, deep abstract, and share.
The primary three assist me resolve what’s value my time, and the fourth means the publish is queued earlier than I transfer on to the following article. Each fixes one thing I used to be tripping over, both in Readwise or in the best way I used to be posting.
Step 1: Triage 10 at a time
Once I open Reader Chomper, it pulls 10 articles from the highest of my “Later” checklist and stops there. If I need extra, I hit Refill, however I’ve to ask for it.
Getting this step proper took longer than any of the particular code did. The issue was Readwise’s default view — it exhibits you all the pieces. I might open the app, see 200 articles, really feel overwhelmed, and shut the tab.
In distinction, 10 articles is sufficiently small to decide to, and it makes the session really feel finite.
I consider it much less as a restrict and extra as a contract with myself: simply learn 10, resolve what to do with each, and I am performed for the day.
Step 2: Run a fast scan
Once I faucet into an article, Reader Chomper provides me a 100-word abstract of three bullets with zero fluff in what I name the Fast Scan.
Most articles finish right here for me, as a result of as soon as I get the three details, I can simply transfer on to the following one. But when the bullets depart me wanting extra, the article will get my full consideration.
The short scan step has made me trustworthy about my studying checklist. I used to persuade myself I might “come again to” articles I knew I would not. With three bullets in entrance of me, the choice takes about 30 seconds: both I’ve bought what I wanted and the article goes to the archive, or there’s one thing right here value chasing.
Step 3: Deep abstract + chat
When the three bullets aren’t sufficient, I’ve two methods to go deeper. The primary is an extended abstract, about 300 phrases as a substitute of 100. The second is what I attain for most frequently: I can chat with the article by means of Claude.
A latest instance: I had an article known as “DuckLake for busy engineering managers” in my queue, which is an information storage strategy that I might by no means heard of. The Fast Scan coated the what, however I nonetheless wasn’t positive if it was related to me. So I requested Claude, “Why is DuckLake related to engineering managers?” Two seconds later, I had a solution with context pulled from the article.
It is a step many read-it-later apps do not (but) have. With it, I can ask my very own questions, and the solutions come straight from the supply.
Step 4: Queue to Buffer
Once I really feel like an article ought to develop into a publish, whether or not after the short scan or an extended chat, I hit Share. Reader Chomper generates a draft: a brief blurb that captures the important thing thought, plus the hyperlink. If the blurb would not fairly learn the best way I might say it, I hit Regenerate, or edit it so it does.
Yet another click on on Queue to Buffer and the publish goes to each my LinkedIn and Bluesky queues.
That is the step I constructed the entire app for. All the things earlier than this level is filtering and deciding what’s value my consideration. The Buffer API handles the final step just about immediately: the publish goes to my queue, and the article will get archived.
The Buffer API was the one piece on this stack I hadn’t labored with earlier than. I wished to share to a number of channels in a single shot, and I additionally wished an excuse to play with it. A number of hundred traces of Swift code later, I had a working integration.
What’s within the stack (and why)
Other than the Buffer API, which I coated in Step 4, the remainder of the stack is deliberately boring. All the things is one thing I used to be already utilizing or already paying for, wired collectively to suit this workflow.
- SwiftUI and native macOS. I thought of a browser extension and an online app, and each would have shipped quicker, however I cared about how the app appeared and the way it carried out. A local app means it launches immediately, scrolls with out lag, and appears prefer it belongs on my Mac. These particulars matter in an app you open day-after-day.
- Readwise Reader. That is the place I save each article I need to learn anyway, and it stays the supply of reality. Reader Chomper is a greater option to get by means of the queue, and something I archive in it will get archived in my Readwise too, maintaining each clear.
- An LLM (Anthropic and OpenAI). I exploit each APIs interchangeably to summarize articles and write the blurbs. The operating spend for the session sits within the nook of the app. It is a small element, nevertheless it jogs my memory the calls aren’t free, and it helps me discover once I’m bouncing between articles for no good purpose.
- SQLite and Dropbox. Articles, summaries, and chat historical past stay in a SQLite file inside my Dropbox folder, which implies the database follows me throughout units with out me having to construct any sync infrastructure of my very own.
What modified (and what did not)
I nonetheless goal for 2 posts per week on LinkedIn and Bluesky, and Buffer queues them up for me. The queue typically runs months deep, so I construct up posts throughout heavy studying weeks and coast on the lighter ones. However I am not posting any greater than I used to be earlier than.
What modified is all the pieces main as much as the publish.
I triage my studying checklist extra actually. Ten at a time forces me to decide, and most articles now finish on the fast scan as a substitute of sitting in a listing unread. I wish to suppose the standard of what I share has modified too: the AI abstract surfaces an thought, and I kind my take from there. Earlier than Reader Chomper, my followers bought a naked hyperlink and needed to click on by means of to determine why they need to care. Now they get a take that does a few of that work for them.
I haven’t got an enormous following, and I am not an excellent poster, however I hope to get higher at it. I am positively nonetheless rising as a creator, however the hole between studying one thing good and posting about it that used to hassle me is now closed. I learn extra thoughtfully now as a result of I do know I’ve a spot to place my opinions, so it’s secure to say my automation is doing its job.
