Today on HackerNoon software updates, we speak to one of our developer's, Jeferson Borba, about his super exciting work with AI-generated TL;DR snippets of HackerNoon stories.
If you haven't yet noticed, as a writer in the story settings there is a new box for TL;DRs, and Jeferson is here to tell us more about it!
This Slogging thread by Limarc Ambalina and Jeferson Borba occurred in slogging's official #software-development channel, and has been edited for readability.
Hey Jeferson Borba! Congrats on this awesome update. So can you explain what is this box? How do writers use it and what does it do?
Hey hey, now we allow authors to add a TL;DR to their stories or let our bot do the hard work for them. Its a nice way give the readers a smaller version of the story. The TL;DR will also be displayed on a new HackerNoon project that will be coming soon 🚀
Awesome tell us a bit about how you made it? Did you draw inspiration from similar tools you’ve already seen out there or is it all from scratch?
Also interested to hear about the AI text generator. How did you build and train that tool?
Oh, I wish I was an AI expert, but actually, this is a tool called TLDRThis. All I had to do was to connect with both our editors and with our back-end, fix some bugs that always appears when using new tools and adapt the calls to our needs
Got it. And what exactly does the tool do? Does it read the entire story and the title and generates a TL;DR based on that or how does it work?
How many words/characters does it generate?
Yep, it reads the entire story and generate a TL;DR based on that. If eventually you thought that you have finished your story, but at a later stage want to return and add more important concepts, you can generate again and your new text will be read by the AI.
I primarily configured it to generate around 500 characters, which is what we suggest for our TL;DR and will fit perfectly for our super secret project that will come out soon.
Great! What problems did you encounter that you weren't expecting? What advice can you give people looking to create a similar integration with TLDRThis using a stack similar to ours? You might be able to save some developers a headache down the line
The TLDRThis is pretty straight forward, my biggest issues were because we have 2 different editors, I had to adapt the use of the API to the different types of outputs that each editor provided. After releasing it, some bugs appeared, mostly because I hadn't tested all the possibilities. Even when we think that we have tested everything, users usually do some unexpected things haha.
If someone is trying to implement something similar to what we have, I wouldn't recommend a TLDR AI other than TLDRThis, is by far the best, super easy to implement and provide human-like TLDR's that sometimes we can't even believe that it was made by an AI.
Very cool update Jeferson Borba! When we can we expect to see the...super cool project where the TL;DRs will be of most use, to launch?! Tease our community!
We are currently in the internal testing phase, I would say that an official version can be released earlier next month, but no promises. It's a pretty complex project and it requires a lot of work. But I really think everyone will enjoy when it's finally released
I'm not pretty sure of how much I can disclose about that project. Anyway, in one of the HackerNoon stories, there is an easter-egg that can give you some clues of what will be coming real soon.
Awesome. Thanks Jeferson looking forward to more of your releases in the future! You're an awesome dev, an awesome person, and we're lucky to have you here.