Posted by seinvak 12/21/2025
I assume this one uses a few-shot LLM approach instead, which is slower and more expensive at inference, but so much faster to build since there's no tedious labeling needed.
Yes, I saw that project pretty impressive! Hand-labeling 4000 books is definitely not an easy task, mad-respect to tracyhenry for the passion and hardwork that was required back then.
For my project, I just used the Gemini 2.5 Flash API (since I had free credits) with the following prompt:
"""You are an expert literary assistant parsing Hacker News comments. Rules: 1. Only extract CLEARLY identifiable books. 2. Ignore generic mentions. 3. Return JSON ARRAY only. 4. If no books found, return []. 5. A score from -10 to 10 where 10 is highly recommended, -10 is very poorly recommended and 0 is neutral. 6. If the author's name is in the comment, include it; otherwise, omit the key. JSON format: [ {{ "title": "book title", "sentiment": "score", "author" : "Name of author if mentioned" }} ] Text: {text}"""
It did the job quite well. It really shows how far AI has come in just 4 years.
These are classics yes, but I was expecting something close to the forefront of the culture
I normally get way better and varied recommendations from my philosophy friends, for example. Here it's generally just the usual mainstream sci-fi stuff about tech, space, ai/robots and such.
And forefront of culture is by definition going to be full of known stuff, else it wouldn't be culture-defining if almost nobody knows it.
What would you put in your top 5 "I'm very smart" 30+ yo book list?
I wonder if it's an american thing, for engineers to be detached of liberal arts? The vibe tends to be quite different in local engineering groups.
The first is that there is likely more diversity the deeper you go down the intellectual hole. You and I may read much more sophisticated books, but the books you read and the ones I read differ significantly. Thus, the list is biased towards the more popular (it is, after all, a popularity list).
Second is this:
> for engineers to be detached of liberal arts?
Most of us just haven't found value in the other types of books. It would help if you gave some examples of books that should be here. For me (perhaps as an engineer), I like books to kind of get to the point. When it comes to fiction, I'm a very firm believer that, although a given novel may give great commentary about a social/philosophical issue, its primary purpose is entertainment. If I wanted to understand the underlying social/philosophical issue, a more direct, nonfiction book will always do a better job.
I've yet to find someone "changed" because of fiction. Those I know who claim to already had the sentiments before they read that piece of fiction, and the story was merely preaching to the choir. What they are glorifying is how well the story depicted an issue.
I think there is a fundamental misunderstanding there, if you think all the potential value of a fiction book is some commentary padded by a story.
Good fiction usually exercises the mind in ways a non-fiction book never would. You experience life through someone else's eyes, you try to understand someone's mind by the actions they take and the words they say, you wonder in a meta-plane what the author is trying to show, you see language being used in non-common ways to provoke emotions or express ideas, you wonder how you would have acted in someone's shoes... Saying the author could get to the point quicker is like saying that lifting weights in the gym is done faster with a forklift, the process is the point rather than the extracted output.
There is also a fundamental difference between being told 'pain is an unpleasant feeling that living beings take effort to avoid' and being punched in the face. Fiction gives you a fraction of the extra wisdom you get with the latter.
>It would help if you gave some examples of books that should be here.
That's the thing, there is not specific book I could recommend that is most likely change your life for the better, for the same reason there is no single specific equation I can mention that will make someone good at math if they solve it. Some exercises are better, some are pointless, but it's the act of engaging that counts in the long term.
My comment about this list had no implication that the books at the top of the list were less valuable than other hidden works; they're just a sign of a path not travelled quite far, if that makes sense.
And leaving aside the usefulness of it all, pleasant experiences not all amount to entertaining. You'd probably agree that having sex with the love of your life and watching TV are not equivalent experiences, even if you come out of both with roughly the same level of self-improvement.
Actually, I'm flipping the two: The potential value of a fiction book is a good story - social commentary is purely optional. Fiction that has commentary padded by a story are valued only by those who are sympathetic to the commentary. Whereas I can easily love a good story even if I disagree with the commentary.
> Good fiction usually exercises the mind in ways a non-fiction book never would. You experience life through someone else's eyes, you try to understand someone's mind by the actions they take and the words they say, you wonder in a meta-plane what the author is trying to show, you see language being used in non-common ways to provoke emotions or express ideas, you wonder how you would have acted in someone's shoes... Saying the author could get to the point quicker is like saying that lifting weights in the gym is done faster with a forklift, the process is the point rather than the extracted output.
I don't think we're in disagreement. I'm merely saying that I've yet to see someone changed by a fiction book. If there was change, it was always "change in the same direction" (e.g. "a renewed appreciation of X").
I have seen plenty of folks changed by nonfiction, though.
Incidentally, most/all of what you wrote above can be done as effectively with nonfiction. Books like When Broken Glass Floats by Chanrithy Him are extremely powerful. As was Killers of the Flower Moon. I doubt any works of fiction dealing with the same topics would be more powerful. Both of these books could have been written (and read) as fiction, but knowing the events were true makes a huge difference in appreciation.
> There is also a fundamental difference between being told 'pain is an unpleasant feeling that living beings take effort to avoid' and being punched in the face. Fiction gives you a fraction of the extra wisdom you get with the latter.
This seems like a false dichotomy. You can have nonfiction do this very effectively without simply "telling" you.
I read a lot, but if I'm going to use a book to make a point or example in a comment, which will be read by someone I don't know, I'll reference a well-known book that most people have heard of, even if it was just from 9th grade English class, instead of something more obscure.
I think you assumed implications in my comment that weren’t at all intended. By all means read whatever you enjoy, and by keeping at it you’ll eventually reach less known work just as a side effect of getting deeper in your niches.
As I said in another comment, I just see the selection as a sign of a lack of exploration.
It’s not that the books themselves are bad -the original super Mario is a nice game, but if it’s the only game being significantly mentioned the crowd is probably not really into gaming. It’s just surprising to me that the HN crowd isn’t really into literature.
One or two might be worthwhile but most would be mainstream and pedestrian.
If you're so well-read why don't you grace us with some of your non-mainstream S-tier recommendations?
From this list, one of the books that I recommend to everyone is Piranesi, which is fairly mainstream, and if they want to explore Russian magical realism then Vita Nostra. Unsong is another favorite. In general I love to explore magic systems that experiment with breaking a system, or stories that explore how different rules might interact within a system.
I think people sometimes underestimate the value of lighter more fun reads, like cultivation stories. The best western adaptation of this style of novel is probably the Cradle series, by Will Wight. Even though the stories tend to be fairly light, they're quite enjoyable for exploring new modes of thinking. For example, we can analyze the interactions of energy as an abstract / symbolic form, and how it influences human behavior; which is an abstract / symbolic application of the cultivation lens over reality. To give an easier to understand example: Feng Shui isn't real but it's true, in the sense that the way in which we organize furniture within a space determines how people navigate it and how they interact. And why might this be useful? Well, sometimes we fail to see the full picture when using a single lens, and different lens might let us see things in a new light.
I've read some terribly generic web novel slop and gotten fairly unique and interesting perspectives from them, but most people aren't good enough readers to enjoy bad books, so they can only read and enjoy good books.
[1] System Programming in Linux:
Playboy is forced to take part in war of the worlds. 50 pages of societé, parties and games are necessary to describe this character.
A number of posts here flagging disambiguation issues, I've run into this a lot.
I've been dealing with the problem using cosine distance between embeddings, but find it tricky to verify at scale.
Anyone else struggling with this?
It seems to miss the mentions of the late John Varley's books in https://news.ycombinator.com/item?id=46269991 six days ago.