Holy crap I have so many thoughts on this having worked on the problem directly. I think it's safe to say that there is a 99.999% chance this will not work, will be launched anyway as a sub-par product, and then killed because it didn't work. And I WORKED in news for eight years. https://twitter.com/Techmeme/status/1339041692679172096

The very TLDR version for this is there is nowhere near enough data to get a high-fidelity bullet-point version. There is *just* enough variation in writer styles that you need to have an algorithm for every writer, and nearly all writers will not have written enough stories!

You can try to abstract it out. Let's say you jump to beats and try to create an algorithm across Apple reporters. Here's one example: create a summarization algorithm that transfers the text of one longform writer to one that often summarizes them. Shocker: not enough data.

Another problem? Writers change beats! They move companies! They carry some style with them, but you lose structural definition and your algorithm gets thrown off. There are multiple inputs that go from writer (unique), to editor (unique), to managing editor, to style guide.

The best you can get really already exists—they're usually called robots. You can get API-level data and write some text around it (stock performance, weather, data like let's say Yelp). But it's Sisyphean to try to capture the essence of a specific story, especially longform.

Have they figured out the secret sauce here? Maybe! Who knows. There are plenty of smart people there. And this was a while ago. But I have a *very* hard time believing that they've cracked this, because the problem is deceptively simple with an innate problem.

You certainly can get bullet points in your news feed. Scraping sentences is easy. But what you will get will basically be a remix of the headline and a lot of low-quality, disconnected information that strips necessary context. At a product level, it's gonna be real bad.