Hacker News new | ask | show | jobs
by jazz3020 623 days ago
OP here—happy to answer any questions! We’ve been working really hard since writing the first line of code 4 months ago.
3 comments

As others already clarified that it is using an LLM do you allow users to report issues with summaries? I would think this might be important.
Yeah, great point. Will add that!
4 months, 73k summaries, 0 mentions of AI on the website/in the FAQ.

How are you doing it?

we used to mention AI in the title, but we removed it. we did list it on a bunch of ai directories like future pedia which does say we use ai

EDIT: just added a help article saying that we use ai

So, AI.

Meaning you “generated” 73k “summaries” of which you are wholly unable to verify the content of because it would require you to have read the books and listened to/read the “summaries”.

What is with sloppybros and thinking quantity is more desirable than quality?

What motivated you to “remove” this fact from your website? Legitimate criticism?

> Meaning you “generated” 73k “summaries” of which you are wholly unable to verify the content of because it would require you to have read the books and listened to/read the “summaries”.

No, we generate summaries on the fly, too. Why would you generate all the summaries in advance when you can summarize real-time when a user clicks on it?

> What is with sloppybros and thinking quantity is more desirable than quality?

Have you read any of our summaries? Go ahead and read a summary of a book you've actually read and let me know what it's lacking. IMO, our summaries are way more thorough than human-written counterparts. And the structure is consistent across the entire catalog. In other words, we offer both quality and quantity.

> What motivated you to “remove” this fact from your website? Legitimate criticism?

No, just a matter of branding. We thought "AI" might add a cool factor, but everyone's AI nowadays, so we're just differentiating. Also, we didn't care to buy the ".ai" domain either. Most people already know it's AI-powered without us letting them know.

> Why would you generate all the summaries in advance when you can summarize real-time when a user clicks on it?

Well the most obvious reason is that the summary page would load faster. I clicked on a book summary and it took much longer to render than even a bloated web app usually takes.

73k summaries isn't that many. If your site gets any traction most of those summaries will be hit repeatedly and will have to generate anyway. Creating the summaries in advance would also allow you to test and verify the output, if you're interested in that.

Okay, that's fair. It isn't that many, but it's still 10x more than the industry leader, Blinkist, though. A good start.

As a small company, we currently need to sacrifice the first reader's experience for the sake of spreading out the costs over time. But every subsequent read is cached!

> No, we generate summaries on the fly, too.

Yikes, so the summaries change with each viewing?

Information that changes each time you read it is by definition “unreliable”.

> Go ahead and read a summary of a book you've actually read and let me know what it's lacking.

Work for a sloppyjoe? And for free?!

Come on. I think you have clearly grown a little too accustom to exploiting other peoples’ work.

Hey, just have your AIs check them all. Right?

It's 2024, it's not worth the time it takes to stop and wonder if it's LLM shovelware. It always is.
true
Which LLM is used for the book summaries and which TTS for the audio?
llm: prefer not to answer tts: open ai
How do you expect anyone to trust the summaries if you are not even going to disclose which AI model generated it?
The most common first action people take on our site is reading a summary of a book they've already read to assess its quality themselves. I don't think they care whether it was written by OpenAI or a monkey, as long as it's good.