Building an AI-powered newsletter Summarizer, Part 1: Why?

Background

I, like any other career-focused engineer, subscribe to an avalanche of engineering newsletters that have quickly overrun my email inbox. The sheer volume of emails i get every day/week has ironically led me to not even try to read them anymore. I just mark each new email as read and swallow the guilt and shame of yet another theoretically good idea ruined by laziness and classic technology overwhelm.

Well, that all ends today.

I’ll be building an LLM-powered summarizer that will serve as a personal newsletter feed that gives me a summary of all the newsletters I receive each week. The reduced bombardment of my inbox should hopefully lead to less information overload and motivate me to actually read those newsletters, jumpstarting my journey to becoming the 10000x engineer I was always destined to be.

How it’ll work

The idea is simple. I set up a local mail server that can receive emails. I change the email addresses on my newsletter subscriptions to route newsletters to the mail server. I configure a weekly cronjob that summarises the newsletters received in the mailbox that week, along with a link to the full texts.

I am admittedly undecided on what the frontend of the summariser should look like. I cannot seem to decide between building a browser extension/web app that receives server sent events and subsequently pushes a desktop notification, or simply sending an email to myself every time the cronjob runs. I am sure the path to follow will become clear to me as I build out the infrastructure.

Motivation

I am lazy, and so will spend considerable amounts of time building a complex technological solution instead of mustering up the discipline to just read my emails.

Not exactly false, but only part of the story.

Truth is, I’ve been holding off on hopping on the AI train for a long time now, mostly for philosophical reasons, but also because I was yet to find a project that wasn’t super contrived and useless. I wanted to build something that would provide me with actual utility, never mind that it was already built by some silicon valley startup with a $200m pre-seed.

This project is equal parts simple and useful to compel me to finally take the plunge into the overcrowded community swimming pool that is the AI space. Let’s hope I don’t lose my underwear in the process :)